Clinical Microbiology and Infectioninfettive.com/wp-content/uploads/2016/08/Diagnosi-C-difficile-ESCMI… · CDI testing. This guideline is intended for use among medical microbiologists,

lable at ScienceDirect

Clinical Microbiology and Infection xxx (2016) 1e19

Contents lists avai

Clinical Microbiology and Infection

journal homepage: www.cl inicalmicrobiologyandinfect ion.com

European Society of Clinical Microbiology and Infectious Diseases:update of the diagnostic guidance document for Clostridiumdifficile infection

M.J.T. Crobach 1, T. Planche 4, C. Eckert 5, F. Barbut 5, E.M. Terveer 1, O.M. Dekkers 2, 3,M.H. Wilcox 6, E.J. Kuijper 1, *

1) Department of Medical Microbiology, Centre for Infectious Diseases, Leiden University Medical Center, Leiden, The Netherlands2) Departments of Clinical Epidemiology and Internal Medicine, Leiden University Medical Center, Leiden, The Netherlands3) Department of Clinical Epidemiology, Aarhus University, Aarhus, Denmark4) Department of Medical Microbiology, St. George's Hospital, London, UK5) National Reference Laboratory for Clostridium difficile, Paris, France6) Department of Microbiology, Leeds Teaching Hospitals & University of Leeds, Leeds, UK

a r t i c l e i n f o

Article history:Received 10 December 2015Received in revised form2 March 2016Accepted 10 March 2016Available online xxx

Editor: D. Raoult

Keywords:Clostridium difficile infectiondiagnosisguidelinerecommendationsreview

* Corresponding author. E. J. Kuijper, Department offor Infectious Diseases, Leiden University Medical CeLeiden, The Netherlands

E-mail address: [email protected] (E.J. Kuijper).

http://dx.doi.org/10.1016/j.cmi.2016.03.0101198-743X/© 2016 The Authors. Published by Elsevierthe CC BY-NC-ND license (http://creativecommons.or

Please cite this article in press as: CrobachMguidance document for Clostridium difficile i

a b s t r a c t

In 2009 the first European Society of Clinical Microbiology and Infectious Diseases (ESCMID) guidelinefor diagnosing Clostridium difficile infection (CDI) was launched. Since then newer tests for diagnosingCDI have become available, especially nucleic acid amplification tests. The main objectives of this updateof the guidance document are to summarize the currently available evidence concerning laboratorydiagnosis of CDI and to formulate and revise recommendations to optimize CDI testing. This update isessential to improve the diagnosis of CDI and to improve uniformity in CDI diagnosis for surveillancepurposes among Europe. An electronic search for literature concerning the laboratory diagnosis of CDIwas performed. Studies evaluating a commercial laboratory test compared to a reference test were alsoincluded in a meta-analysis. The commercial tests that were evaluated included enzyme immunoassays(EIAs) detecting glutamate dehydrogenase, EIAs detecting toxins A and B and nucleic acid amplificationtests. Recommendations were formulated by an executive committee, and the strength of recommen-dations and quality of evidence were graded using the Grades of Recommendation Assessment, Devel-opment and Evaluation (GRADE) system. No single commercial test can be used as a stand-alone test fordiagnosing CDI as a result of inadequate positive predictive values at low CDI prevalence. Therefore, theuse of a two-step algorithm is recommended. Samples without free toxin detected by toxins A and B EIAbut with positive glutamate dehydrogenase EIA, nucleic acid amplification test or toxigenic culture re-sults need clinical evaluation to discern CDI from asymptomatic carriage. M.J.T. Crobach, CMI 2016;▪:1© 2016 The Authors. Published by Elsevier Ltd on behalf of European Society of Clinical Microbiology and

Infectious Diseases. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Introduction

The previous European Society of Clinical Microbiology and In-fectious Diseases (ESCMID) guidance document for Clostridiumdifficile infection (CDI) was published in 2009 [1]. Since then manylaboratories in Europe have implemented a diagnostic algorithmfor diagnosing CDI. However, many new diagnostic tests have

Medical Microbiology, Centrentre, PO Box 9600, 2300 RC

Ltd on behalf of European Societyg/licenses/by-nc-nd/4.0/).

JT, et al., European Society of Cnfection, Clinical Microbiolog

become available in the meantime, especially nucleic acid ampli-fication tests (NAATs). Although several of these tests have beenmarketed, their role in the diagnosis of CDI needs to be clarified.Also, the importance of free toxin detection in stool needs to beaddressed. This update of the previous guidance document isessential to improve the diagnosis of CDI; to optimize its manage-ment, prevention and control; and to improve uniformity in CDIdiagnosis for surveillance purposes across Europe.

The main objectives of this guidance document are to sum-marize the currently available evidence concerning laboratorydiagnosis of CDI and to formulate recommendations to optimizeCDI testing. This guideline is intended for use among medicalmicrobiologists, gastroenterologists, infectious disease specialists

of Clinical Microbiology and Infectious Diseases. This is an open access article under

linical Microbiology and Infectious Diseases: update of the diagnosticy and Infection (2016), http://dx.doi.org/10.1016/j.cmi.2016.03.010

http://creativecommons.org/licenses/by-nc-nd/4.0/


mailto:[email protected]

www.sciencedirect.com/science/journal/1198743X

http://www.clinicalmicrobiologyandinfection.com

http://dx.doi.org/10.1016/j.cmi.2016.03.010




M.J.T. Crobach et al. / Clinical Microbiology and Infection xxx (2016) 1e192

and infection control practitioners. The target population is diar-rhoeal patients suspected of having CDI.

Material and Methods

To be able to revise our previous recommendations, an update ofthe 2009 meta-analysis was performed. In addition, other guide-lines and recent literature concerning the diagnosis of CDI werereviewed.

Update of meta-analysis

Search strategyStudies evaluating laboratory assays for diagnosing CDI were

searched in PubMed, Embase, Web of Science, Central and theCochrane Library. Searches were performed in June 2014 with thesupport of a trained librarian. The search was restricted to articlespublished since 2009 in the English language. Meeting abstractswere excluded. The search strategy is displayed in SupplementaryMaterial 1.

Table 1The 2 � 2 table used to calculate test characteristics

Diseased or referencetest positive

Not diseased orreference test negative

Index test positive (a) True positive (b) False positiveIndex test negative (c) False negative (d) True negative

Reference testsA reference test is the best available test and is the standard

against which other assays are compared. Cell cytotoxicityneutralization assay (CCNA) and toxigenic culture (TC) are regardedas reference tests for diagnosing CDI [2].

CCNA demonstrates the presence of free toxin B. For this test,stool filtrates are inoculated onto a monolayer of a cell culturewhich is then observed for a toxin B-induced cytopathic effect(rounding of the cells). The cytopathic effect is evaluated at 24 and48 hours. Cell lines commonly used for CCNA include Vero cells,HeLa cells, human foreskin fibroblast cells and Hep-2 cells.Neutralization of the cytopathic effect is necessary to determinethe specificity of this effect and can be done by using Clostridiumsordelli antitoxin or C. difficile antitoxin [3]. This reference testtakes 1 to 2 days to perform and requires cell culture and labo-ratory expertise, so it is not routinely used in most diagnosticlaboratories.

TC demonstrates the presence of C. difficile, which is able toproduce toxins in vitro. Stools are incubated anaerobically for atleast 48 hours on selective media. Many different culture mediaexist for this purpose, all aiming to enhance the recovery ofC. difficilewhile inhibiting the overgrowth of other faecal flora [4].Pretreatment with alcohol shock [5] or heat shock can also beused to decrease overgrowth of normal faecal flora [4]. Also, brothenrichment before plating onto a solid medium is sometimes used(also called enriched culture) [4]. Furthermore, a chromogenicmedium (ChromID agar; bioM�erieux) for the recovery of C. difficilehas been developed which is designed to isolate and identifyC. difficile within 24 hours. However, no consensus exists onwhich culture medium and/or culture method is the mostappropriate to use. Colonies suspicious for C. difficile can berecognized by Gram staining, colony morphology, ‘horse manure’odour, biochemical testing, gaseliquid chromatography, ultravi-olet light fluorescence, latex agglutination and matrix-assisteddesorption ionizationetime of flight mass spectrometry [6]. Iso-lates from positive cultures are either tested for in vitro toxinproduction by the use of CCNA or toxin A/B enzyme immunoassay(EIA) or tested for the presence of toxin A/B genes by NAAT.

As outlined above, both these reference tests detect differentthings, and because of this theywill not necessarily agree with eachother in all samples. Results for each reference test will be analysedseparately.

Please cite this article in press as: CrobachMJT, et al., European Society of Cguidance document for Clostridium difficile infection, Clinical Microbiolog

Index testsIndex tests are the tests whose performance is being evaluated

compared to the reference tests. The index tests we reviewedcomprise all commonly applied and commercially available labo-ratory tests for diagnosing CDI other than the reference tests. Theseinclude EIAs that detect glutamate dehydrogenase (GDH), EIAs thatdetect toxins A and B and NAAT.

GDH EIAs detect glutamate dehydrogenase, an enzyme that isproduced by both toxigenic and nontoxigenic strains of C. difficile.GDH EIAs are available in well-type format (results are displayed asa colour change which can be detected visually or photo-spectrometically) ormembrane-type format (results can be visuallyread from a membrane).

Toxin A/B EIAs detect toxins A and B and are also available inwell-type or membrane-type format. Most EIAs detecting onlytoxin A have been replaced by EIAs detecting both toxins A and B, asstrains that only produce toxin B and not toxin A are reported.

Several membrane-type tests that include both an EIA detectingGDH and an EIA detecting toxins A and B are also available (C. diffQuik Chek Complete, Techlab, Combo C. difficile; Theradiag).

NAATs include assays that use PCR, helicase-dependentamplification and loop-mediated isothermal amplification. Mostassays detect conserved regions within the gene for toxin B (tcdB),but assays that detect a highly conserved sequence of the toxin Agene (tcdA) have also been developed (Illumigene, Meridian,Bioscience and Amplivue, Quidel) [7,8]. NAATs that not only detecttcdB but also the binary toxin genes (cdt) and the deletion atnucleotide 117 on tcdC are also available (Verigene C. difficile test,Nanosphere and Xpert, Cepheid) and offer the potential advantageof detecting PCR ribotype 027, although highly related PCR ribo-types may also be detected by these tests (without distinguishingthem from PCR ribotype 027) [9]. NAATs that detect multipletargets at the same time, including C. difficile toxin genes, are alsoavailable (Seeplex Diarrhea ACE detection, Seegene, xTAGGastrointestinal Pathogen Panel, Luminex, FilmArray Gastroin-testinal Panel, BioFire Diagnostics).

Test performanceThe numbers of truly positive, falsely positive, falsely negative

and truly negative index test results are generally displayed in a2� 2 table (Table 1). Test performance can be derived from this 2� 2table. The sensitivity of a test is defined as the probability that theindex test result will be positive in a person with disease (a/aþ c).The specificity of a test is defined as the probability that the indextest result will be negative in a personwithout disease (d/bþ d). Thepositive predictive value (PPV) of a test is the probability that aperson has the disease, given the positive test result (a/aþ b). Thenegative predictive value (NPV) of a test is the probability that aperson is free of disease, given the negative test result (d/cþ d). PPVand NPV are dependent on disease prevalence in the tested popu-lation (http://training-old.cochrane.org/sites/training-old.cochrane.org/files/uploads/DTA/1.3_Introduction_to_test_accuracy/story.html).

Eligibility criteriaStudies eligible for inclusion had to: (1) describe original

research, (2) compare an index test (one commercially available in


http://training-old.cochrane.org/sites/training-old.cochrane.org/files/uploads/DTA/1.3_Introduction_to_test_accuracy/story.html



M.J.T. Crobach et al. / Clinical Microbiology and Infection xxx (2016) 1e19 3

Europe) with a reference test (CCNA or TC), (3) perform the tests onC. difficile-negative and -positive clinical human stool samples and(4) provide sufficient information to recalculate sensitivity andspecificity and their confidence intervals. Culture without deter-mining the toxigenic status was accepted as a reference test if onlyassays detecting GDH were evaluated.

Studies were excluded if: (1) the reference test was not per-formed on all samples but only on positive, negative or discordantsamples (to exclude partial verification bias), (2) not all sampleswere tested by the same reference test, (3) the reference methodwas a composite of more than one test, (4) the reference methodincluded clinical data for its interpretation, (5) the index test waspartly used as reference method, (6) the index test did not followmanufacturers' instructions for testing or sample collection, (7) forCCNA, samples were not stored correctly before testing (refriger-ated or frozen at �20�C and thawed only once) or neutralization todetermine the specificity of the cytopathic effect was not executedand (8) only selected samples were included.

Selection processStudy eligibility was assessed in a two-step selection process by

two independent investigators (MC, ET). Inconsistencies wereresolved by consensus and by consultation of a third and fourthinvestigator (EK, TP).

Outcome measures, data extraction and quality assessmentThe principal measures of outcome were the sensitivity and

specificity of different index tests compared to one of the 2 refer-ence tests. Toxin A/B EIAs, GDH EIAs and NAATs were compared toCCNA and TC. GDH EIAs were additionally compared to culture.From each study we extracted the number of true-positive, false-positive, false-negative and true-negative findings to be able tocalculate the sensitivity and specificity of the index test evaluatedin that study. Data were extracted by two independent in-vestigators (MC, ET) using a data extraction form (SupplementaryMaterial 2). Additional data that were extracted included year ofpublication, storage conditions of the samples, information aboutthe study population and information about the execution of theindex test and reference test.

The quality of the studies was assessed by the same two inde-pendent investigators using a quality assessment tool. This qualityassessment tool (Supplementary Material 3) consisted of itemsfrom the Quality Assessment for Studies of Diagnostic Accuracy(QUADAS) tool [10], supplemented with items concerning theappropriate handling of specimens and appropriate execution ofreference tests.

Statistical analysisFor all index tests in all studies, the sensitivity and specificity

and their respective confidence intervals were calculated from the

Table 2Scoring system for grading quality of evidence and strength of recommendations

Quality of evidenceHigh quality Evidence from at least one properly desig

and direct comparison of all test results wModerate quality Evidence from: (1) at least one cross-sec

results with an appropriate reference staLow quality Evidence from opinions of respected auth

or reports of expert committees.Strength of recommendationStrong recommendation for use Desirable effects clearly outweigh undesiWeak recommendation for use Desirable and undesirable effects are closWeak recommendation against use Desirable and undesirable effects are closStrong recommendation against use Undesirable effects clearly outweigh desiGood practice statement Desirable effects clearly outweigh undesi


number of true-positive, false-positive, false-negative and true-negative findings supplied in these studies. Wherever possible,the results after initial testing (instead of results after retesting ofindeterminate results) were used to calculate the sensitivity andspecificity. Random effects logistic regression was used to pool themean sensitivities and specificities for the different index tests andthe different types of index tests. In case of fewer than four studies,a fixed effect model was used. NPVs and PPVs were calculated usinga hypothetical prevalence of CDI of 5, 10, 20 and 50% in the testedpopulation. We used Stata 12.0 software (StataCorp) for all statis-tical analyses.

Guidelines and additional studies

An electronic search was performed on topics concerning lab-oratory diagnosis of CDI not included in our meta-analysis (e.g.repeated testing, sample selection). Published guidelines on CDItesting were also studied. These included guidelines from the So-ciety for Healthcare Epidemiology of America/Infectious DiseasesSociety of America (published in 2010) [11], guidelines from theAustralasian Society for Infectious Diseases (published in 2011)[12], guidelines from the American College of Gastroenterology(published in 2013) [13], guidelines from the American Academy ofPediatrics (published in 2013) [14] and guidelines from the UKNational Health Service (update published in 2012) [15].

Formulation of recommendations

The guideline was developed according to the Appraisal ofGuidelines for Research and Evaluation (AGREE II) instrument [16].Findings of the literature review and meta-analysis results werediscussed with the members of the executive committee, and rec-ommendations were formulated. We slightly modified the GRADEsystem to grade the strength of the recommendations and thequality of evidence [17] (Table 2). A good practice statement couldbe made instead of a formal graded recommendation for domainswhere this was deemed appropriate [18]. The drafting group (con-sisting of experts in the field) and a patients' representative wereinvited to comment on the recommendations, and results fromthese discussions were incorporated in the final recommendations.

Results

Literature search and selection process

A total of 795 unique citations were identified by our currentsearch. On the basis of title and abstract, 693 articles were excluded,leaving 102 full-text articles for detailed assessment. In total, 61studies were excluded after detailed assessment. Reasons forexclusion were (some studies had more than one reason for

ned cross-sectional or cohort study in patients with diagnostic uncertaintyith an appropriate reference standard.

tional or cohort study in selected patients and/or no or partial comparison of testndard, (2) caseecontrol studies.orities, based on clinical experience, descriptive case studies

rable effects.ely balanced or recommendation is based on low-quality evidence.ely balanced or recommendation is based on low-quality evidence.rable effects.rable effects, but no or only indirect evidence is/will become available.



exclusion): not all samples were tested by the (same) referencemethod (23 studies), no or an inadequate reference test was used(16 studies), samples were selected inadequately (13 studies), notenough information was provided (seven studies), the study didnot describe original research (five studies), no clinical human stoolsamples were included (three studies), no commercial diagnostictest was investigated (two studies) and stool samples were incor-rectly collected in transport medium (one study).

From all 43 studies included in the previous meta-analysis [1],28 were excluded. Twenty-four of these studies evaluated tests thatwere no longer available (mainly EIAs detecting toxin A only). Twoother studies were excluded because they did not evaluate acommercial test (both studies evaluated an in-house PCR), onestudy was excluded because not all samples were tested by thesame reference test and one study was excluded because sampleswere stored incorrectly for CCNA testing. A total of 56 studies (15from the previous meta-analysis and 41 published since 2009)were included in the meta-analysis [7,8,19e72]. A summary of theselection process is shown in Fig. 1.

Study characteristics

Twenty-four different laboratory assays were evaluated: onewell-type EIA for GDH, three membrane-type EIAs for GDH, five

43 studies from the previous review

102 studies selected for detailed assessment

795 potentially relevant studies identified by literature search

(published since 2009)

56 studies included in current meta-analysis

41 studies included

15 studies included

Fig. 1. Summary of selection process. *Some stud


well-type EIAs for toxins A and B, four membrane-type EIAs fortoxin A and B and 11 NAATs (Table 3). In total, 133 comparisonsbetween index tests and reference tests were available, including53 comparisons to CCNA, 69 comparisons to TC and 11 comparisonsto culture. Studies were published between 1996 and 2014. Thenumber of evaluated index tests per study ranged from one to ten,and the number of included samples ranged from 60 to 12 369. TheCDI prevalence in the tested population ranged from 6 to 48%.Table 4 lists the characteristics of included studies.

Quality assessment

None of the studies fulfilled all our quality assessment criteria,mainly because required information was frequently missing(Fig. 2, Supplementary Material 4). The process used to selectsamples was adequately reported in 23 (41%) of 56 studies. Aminority of studies (6/56, 11%) reported that they did not excludeformed samples from CDI testing. In around half of the studies,conditions of storage for the samples before testing with theindex test were not (or were insufficiently) reported. Samplestested by GDH EIA, toxin A/B EIA and NAAT were reported to bestored according to manufacturer's instructions in 10 (46%) of 22,14 (45%) of 31 and 15 (50%) of 30 studies, respectively. In theremaining 12, 16 and 15 studies, respectively, storage conditions

28 articles excluded:24 test not available anymore2 no commercial test1 not all samples tested by same reference test1 samples for CCNA were stored incorrectly

693 studies excluded on the basis of title or abstract

61 articles excluded*:23 not all samples tested by (same) reference test16 no or incorrect reference test13 incorrect sample selection7 not enough information5 no original research3 no clinical human stool samples2 no commercial diagnostic test1 incorrect sample collection

ies had more than one reason for exclusion.


Table 3Index tests included in meta-analysis

Assay type Test Manufacturer Target Method

(A) Well-type EIA GDH C. diff Chek-60 Techlab GDH Well-type EIA(B) Membrane-type EIA GDH C. diff Quik Chek Techlab GDH Membrane-type EIA

ImmunoCard C. difficile Meridian GDH Membrane-type EIAQuik Chek CompleteeGDHa Techlab GDH Membrane-type EIA

(C) Well-type EIA toxins A/B Premier toxins A/B Meridian Toxins A and B Well-type EIARemel ProSpecT Oxoid Toxins A and B Well-type EIARidascreen toxins A/B Biopharm Toxins A and B Well-type EIAClostridium difficile Tox A/B II Techlab Toxins A and B Well-type EIAVidas CDAB bioM�erieux Toxins A and B Automated EIA

(D) Membrane-typeEIA toxins A/B

ImmunoCard toxins A/B Meridian Toxins A and B Membrane-type EIAQuik Chek CompleteeTox A/Ba Techlab Toxins A and B Membrane-type EIATox A/B Quik Chek Techlab Toxins A and B Membrane-type EIAXpect Oxoid Toxins A and B Membrane-type EIA

(E) NAAT Advansure CD LG Life Sciences tcdA, tcdB RT-PCRAmplivue Quidel tcdA Isothermal helicase-dependent

amplificationBD GeneOhm Becton Dickinson tcdB RT-PCRBD Max Cdiff Becton, Dickinson tcdB RT-PCRGenomEra Abacus Diagnostica tcdB RT-PCRIllumigene Meridian tcdA LAMPPortrait Great Basin tcdB Isothermal helicase-dependent

amplificationProdesse ProGastro Cd Assay Hologic Gen-Probe tcdB RT-PCRSeeplex Diarrhea ACE Detectionc Seegene tcdB RT-PCRVerigene Nanosphere tcdA, tcdB, cdt,b tcdC deletion nt 117b PCR/nanoparticle-based microarrayXpert C. difficile Cepheid tcdB, cdt, tcdC deletion nt 117 RT-PCR

EIA, enzyme immunoassay; GDH, glutamate dehydrogenase; LAMP, loop-mediated isothermal DNA amplification; RT-PCR, real-time PCR.a Part of an EIA that detects both toxins A/B and GDH.b Only for epidemiologic purposes.c Multiplex PCR system.


did not or not completely comply with manufacturer's in-structions. In 18 (72%) of 25 studies using CCNA as the referencetest, samples were stored according to our predefined storagerequirements: samples were either refrigerated and testedwithin 5 days (15 studies) [8,25,27,36,45e48,58e61,63,65,68]or were frozen at �20�C and thawed no more than once (threestudies) [44,66,67]. In the remaining seven studies (28%), storageconditions for CCNA were not or incompletely described.Storage conditions for samples tested by TC were reported in 23(68%) of 34 studies, but no specific requirements for storage ofsamples tested by TC were set. The execution of the referencetest was described in sufficient detail in 44 (79%) of 56 studies. In2 (8%) of 26 studies using CCNA as reference test, theincubation period was only 24 hours [61,63]. In studies using TCas reference test, ethanol shock was reported to be performedin 18 of 35 studies [19,21,23,32,35,37,38,47,51e55,57,61,69e71],and heat shock was performed in three of 35 studies [22,49,58].Eight studies (23%) used an enrichment broth before platingonto a solid agar [19,22e24,32,43,58,62]. Toxigenicity wasconfirmed by PCR (15/32, 47%) [21,23,29,33e35,37,51e57,70],CCNA (9/32, 28%) [7,8,22,24,43,47,58,61,62], toxin EIA (7/32, 22%)[19,30,32,38,40,69,71] or both PCR and CCNA (1/32, 3%) [26].Blinding (index test interpreted without knowledge of referencetest or vice versa) was reported in 8 (14%) of 56 studies. Thirty-one studies (55%) reported if any indeterminate results (i.e.invalid, ‘no call’ or difficult-to-interpret results) were found.Indeterminate results actually occurred in 28 studies and werereported for one membrane-type GDH EIA (ImmunoCardC. difficile), three membrane-type toxin A/B EIAs (Tox A/B QuikChek, ImmunoCard Tox A/B, Xpect), one automated EIA (Vidas)and nine NAATs. The amount of indeterminate results rangedfrom 0.3 to 6.8% of tested samples. Repeat testing of samplesafter an initial indeterminate result was done in 24 (86%) of these28 studies. Of these, 22 presented results only after repeat testing[7,8,20e22,24,29,30,34,35,37,38,43,46,47,54,58,59,62,65,69,70],


and two presented results of both initial and repeat testing[27,63].

Test performances

Sensitivity and specificity of the index tests were calculated onthe basis of the numbers provided in the articles. Discrepanciesbetween calculated sensitivity or specificity and published datawere found in two articles; the correct datawere provided by bothauthors upon request [38,39]. In Table 5, sensitivity and specificityof index tests are compared to CCNA. Reported estimates ofsensitivity ranged from 0.80 to 1.00 for GDH EIAs, from 0.44 to0.99 for toxin A/B EIAs and from 0.83 to 1.00 for NAATs. Reportedestimates of specificity ranged from 0.82 to 0.95 for GDH EIAs,from 0.87 to 1.00 for toxin A/B EIAs and from 0.87 to 0.98 forNAATs. Table 6 lists sensitivity and specificity compared to TC.Sensitivities ranged from 0.83 to 1.00, 0.29 to 0.86 and 0.77 to 1.0for GDH EIAs, toxin A/B EIAs and NAATs, respectively. Specificitiesranged from 0.88 to 1.00, 0.91 to 1.00 and 0.83 to 1.00, respec-tively. In Table 7, sensitivity and specificity of GDH EIAs arecompared to culture. Sensitivities ranged from 0.71 to 1.00, andspecificities ranged from 0.67 to 1.00. In Table 8, estimates ofpooled sensitivity and pooled specificity for the different cate-gories of index tests are shown. The estimated pooled sensitivitiesand specificities compared to CCNA were used to compute PPVsand NPVs of the categories of index tests at different hypotheticalCDI prevalences (Table 9, Supplementary Material 5). At a CDIprevalence of 5%, PPVs ranged from 34 to 81%, and NPVs rangedfrom 99 to 100%. At a CDI prevalence of 50%, PPVs ranged from 91to 99%, while NPVs ranged from 83 to 98%.

Discussion

In the present meta-analysis, we evaluated the diagnostic ac-curacy of various commercial laboratory assays for diagnosing CDI.


Table 4Characteristics of included studies

Study Year Country Reference test Index test Total no. samples Study population Consistency ofstool samples

Prevalence CDI(CCNA)

Prevalence CDI(TC)

Barkin [19] 2012 USA TC Premier toxins A/B,ImmunoCard C. difficile,Illumigene

272 Adult inpatients of large community teachinghospital with diarrhoea, risk factors for CDI andfor whom CDI test was requested by theirphysician

Unformed 13.1

Berg, van den [66] 2005 Netherlands CCNA ImmunoCard toxins A/B 367 Unformed stools of adults with specific requestfor CDI testing or hospitalized >72 hours thatwere submitted to laboratories of threeuniversity hospitals

Unformed 6.3

Berg, van den [67] 2007 Netherlands CCNA Premier toxins A/B 540 Unformed stools of patients suspected of havingCDI or hospitalized >72 hours in four universitymedical centres

Unformed 5.7

Berry [20] 2014 UK CCNA Xpert 1034 Inpatients in two acute-care hospitals aged>15 years with suspected CDI for whom CDItesting was requested by treating physician

Unformed 6.0

Boer, de [25] 2010 Netherlands CCNA Xpect 161 Clinical stool specimens from patients forwhom request for CDI testing was issued,prospectively collected at laboratory forinfectious diseases

Unclear 9.9

Bruins [21] 2012 Netherlands TC ImmunoCard toxins A/B,Quik Chek Complete,Premier toxins A/B,Illumigene

986 Hospitalized and nonhospitalized patients withdiarrhoea who had stool sample sent tolaboratory of major hospital, preferably fromthose patients known to have CDI-associatedsymptoms or risk factors

Unformed 7.4

Buchan [22] 2012 USA TC Portrait, GeneXpert,GeneOhm, Illumigene

540/275/169/96 Stool specimens from patients >2 years oldsuspected of having CDI collected at fourinstitutions

Unformed 22.5

Calderaro [23] 2013 Italy TC Illumigene, Quik ChekComplete

306 Patients attending university hospital withsuspicion of CDI

Unclear 19.6

Carroll [24] 2013 USA TC Verigene 1875 Leftover stool samples submitted specifically forCDI testing according to institution's routinepractice to five geographically diverse clinicalmicrobiology laboratories

Formed andunformed

8.4 (direct),14.7 (enriched)

Eastwood [27] 2009 UK CCNA Premier toxins A/B, Xpect,Tox A/B Quik Chek,Ridascreen toxins A/B, ToxA/B II, ProSpecT, VIDASCDAB, ImmunoCard toxinsA/B, C. diff Chek-60, BDGeneOhm

488 Stool specimens submitted for CCNA testing atlaboratory of teaching hospital; ten sampleswere randomly chosen each day

Unformed 18.1

M.J.T.Crobach

etal./

ClinicalMicrobiology

andInfection

xxx(2016)

1e19

6Pleasecite

thisarticle

inpress

as:CrobachMJT,etal.,European

SocietyofClinicalM

icrobiologyand

InfectiousDiseases:update

ofthediagnostic

guidancedocum

entfor

Clostridiumdifficile

infection,ClinicalMicrobiology

andInfection

(2016),http://dx.doi.org/10.1016/j.cmi.2016.03.010

Eckert [8] 2014 France CCNA, TC Amplivue, C. diff Quik Chek 308 Inpatients in four university-affiliated hospitals>2 years old with suspected CDI for whom CDItesting was requested by treating physician or ifdiarrhoea occurred after day 3 of hospitalization

Unformed 7.5 11.7

Fenner [28] 2008 Switzerland Culture C. diff Chek-60 1468 Stools of adults patients suspected of having CDIat university hospital

Unclear 12.7 culturepositive

Hart [29] 2014 Australia Culture, TC Illumigene, BD GeneOhm,Quik Chek Complete

150 Stools of children collected at laboratory ofpaediatric hospital fulfilling criteria for CDItesting in this hospitala

Formed (4%)and unformed(96%)

30.0

Hirvonen [30] 2013 Finland TC GenomEra 310 Stool specimens from inpatients (7e95 yearsold), collected prospectively according toroutine hospital practice for antibiotic-associated diarrhoea at large teaching hospital

Unformed 24.9

Huang [31] 2009 Sweden CCNA Xpert 220 Consecutive stool specimens from patients>2 years old who were symptomatic and hadrequest for CDI testing at university hospital

Unformed 10.5

Jacobs [32] 1996 Israel Culture, TC ImmunoCard C. difficile 258 Stool samples from patients who developeddiarrhoea during hospitalization in communityteaching hospital and control samples from 24patients without diarrhoea

Formed andunformed

7.0

Jong [26] 2012 Netherlands TC ImmunoCard toxins A/B,VIDAS CDAB

150 Hospitalized adult patients in tertiary teachinghospital who had stool specimens submitted forCDI testing

Unclear 9.7

Kawada [33] 2011 Japan Culture, TC Quik Chek Complete,ImmunoCard C. difficile, ToxA/B Quik Chek

60 Patients hospitalized at geriatric hospital anddiagnosed as having antibiotic-associateddiarrhoea

Unformed 46.7

Kim [35] 2014 Korea TC Quik Chek Complete, VIDASCDAB

608 Suspected CDI patients in tertiary-care teachinghospital

Unformed 9.0

Kim [34] 2012 Korea TC AdvanSure, VIDAS CDAB 127 Diarrhoeal stool specimens submitted tohospital laboratory for C. difficile culture

Unformed 8.8

Lalande [7] 2011 France TC Illumigene 472 Consecutive stools from patients suspected ofhaving CDI

Unformed 10.4

Larson [36] 2010 USA CCNA C. diff Quik Chek 699 Stool samples submitted for CDI testing fromadult patients at university hospital

Unformed 6.7

Le Guern [37] 2012 France TC BDMax Cdiff, BD GeneOhm,Tox A/B Quik Chek

360 Diarrhoeal stool specimens collected frominpatients at university hospital

Unformed 12.2

(continued on next page)

M.J.T.Crobach

etal./


andInfection

xxx(2016)

1e19

7

Pleasecite

thisarticle

inpress


SocietyofClinicalM

icrobiologyand


ofthediagnostic

guidancedocum

entfor



andInfection


Table 4 (continued )



Prevalence CDI(TC)

Leitner [38] 2013 Austria TC BD Max Cdiff, Premiertoxins A/B

180 Stool specimens from adults and children withspecified request for CDI testing at medicaluniversity

Unformed 16.7

Massey [39] 2003 Candada CCNA Tox A/B II 557 Stool samples of adult hospitalized patientssuspected of having CDI at large teachinghospital

Unformed 25.7

Mattner [40] 2012 Germany TC Ridascreen 254 All liquid stool samples sent to universitymicrobiology laboratory

Unformed 16.4

Musher [41] 2007 USA CCNA Premier toxins A/B,ImmunoCard toxins A/B,Tox A/B II, ProSpecT

446/131 Consecutive stool samples submitted tolaboratory of medical centre for CDI testing

Unclear 17.0/41.2

Noren [42] 2011 Sweden CCNA Illumigene 272 Consecutive stool specimens from adults andchildren submitted for CDI testing fromhospitals and communities

Unclear 13.2

Novak-Weekley[43]

2010 USA TC Xpert, Premier A/B 432 Leftover stool samples from patients >2 yearsold with suspected CDI for whom toxin enzymeimmunoassays were ordered according toinstitution's standard practices at regionalreference laboratories serving hospitals andassociated medical clinics

Unformed 16.8

O'Connor [44] 2001 Ireland CCNA Tox A/B II, Premier toxins A/B

200 Consecutive stools of adult patients suspectedof having CDI submitted to laboratories ofuniversity hospitals

Formed andunformed

30.5

Ota [45] 2012 USA CCNA C. diff Quik Chek Complete,Premier toxins A/B,Illumigene

141 Consecutive stool specimens prospectivelycollected at children's hospital from patients 1e18 years of age and submitted for CDI testing

Unformed 18.4

Pancholi [46] 2012 USA CCNA Illumigene, Xpert 200 Consecutive and prospectively collected stoolsfrom adult patients submitted to universitymedical centre laboratory for routine CDItesting

Unformed 11.6

Planche [47] 2013 UK CCNA, TC Xpert, C. diff Chek-60,Premier toxins A/B, Tox A/BII

8827/12 365/9192/12 369

Faecal samples from hospital and communitypatients submitted for routine CDI testingaccording to routine protocolb submitted to fourhospital diagnostic laboratories serving majorteaching hospitals and their communities

Unformed 5.9 8.4

Qutub [48] 2011 Saudi Arabia CCNA C. diff Chek-60, Tox A/B II 150 Stool samples from consecutive inpatients withsuspected CDI

Unformed 34.7

M.J.T.Crobach

etal./


andInfection

xxx(2016)

1e19

8Pleasecite

thisarticle

inpress


SocietyofClinicalM

icrobiologyand


ofthediagnostic

guidancedocum

entfor



andInfection


Reller [49] 2007 USA Culture C. diff Chek-60 439 Stool samples from hospitalized adults andchildren suspected of having CDI

Unclear 36.7 culturepositive

Reller [50] 2010 USA CCNA C. diff Chek-60, C. diff QuikChek, Tox A/B Quik Chek

600 Sequential weekday stool samples submitted touniversity hospital microbiology laboratory forCDI testing

Unformed 7.7

Shin [52] 2009 Korea TC Vidas CDAB 1596 Stool samples from patients admitted to tertiaryteaching hospital with clinical signs compatiblewith CDI

19.6

Shin [51] 2009 Korea TC Vidas CDAB 555 Patients >2 years old with suspected CDI fromtwo hospitals

Formed (51%)and unformed

20.3

Shin [53] 2012 Korea TC Seegene, BD GeneOhm 243 Fresh stool specimens from patients withclinical signs compatible with CDI who werehospitalized in 3 teaching hospitals

Unclear 28.8

Shin [54] 2012 Korea TC Xpert/epi, Vidas CDAB 253 Consecutive stool specimens from suspectedCDI patients in tertiary hospital

Unformed 18.4

Sloan [55] 2008 USA TC Premier toxins A/B, Xpect,ImmunoCard A/B

200 Stools of patients suspected of having CDIsubmitted to clinical microbiology laboratory oflarge tertiary-care teaching hospital

Unformed 22.0

Snell [56] 2004 Canada Culture, TC C. diff Chek-60, Tox A/B II 497 Stools of inpatients suspected of having CDI atlarge teaching hospital

Unformed 10.5

Soh [57] 2014 Korea TC AdvanSure CD, Illumigene 203 Stool samples collected at tertiary universityteaching hospital

Unformed 12.8

Stamper [59] 2009 USA CCNA BD GeneOhm 401 Symptomatic adult patients who had stoolsample submitted for routine CDI testing intertiary-care university medical centre

Unformed 11.0

Stamper [58] 2009 USA CCNA, TC ProGastro CD 280 Stool samples submitted for routine CDI testingfrom symptomatic patients >2 years old attertiary-care university medical institution

Unformed 11.0 15.7

Staneck [60] 1996 USA CCNA ImmunoCard C. difficile 906 Stool samples submitted to three hospitalmicrobiology laboratories

Unclear 14.1

Swindells [61] 2010 UK Culture, CCNA,TC

C. diff Quik Chek Complete,Vidas CDAB, Xpert,GeneOhm

150 Consecutive stool specimens from inpatients>65 years old who developed diarrhoea at least48 hours after admission

Unformed 10.0 12.0

Tenover [62] 2010 USA/Canada TC Xpert 2296 Leftover stool specimens from patients >2 yearsold from seven health care organizations (sixUSA, one Canada) for whom CDI testing wasordered according to institution's practices

Unformed 10.8 (direct),14.7 (enriched)

(continued on next page)

M.J.T.Crobach

etal./


andInfection

xxx(2016)

1e19

9

Pleasecite

thisarticle

inpress


SocietyofClinicalM

icrobiologyand


ofthediagnostic

guidancedocum

entfor



andInfection


Table 4 (continued )



Prevalence CDI(TC)

Terhes [63] 2009 Hungary CCNA BD GeneOhm 600 Inpatients and outpatients at local universityhospital who had diarrhoeal stool sample sentto laboratory for CDI testing

Unformed 6.4

Ticehurst [64] 2006 USA CCNA C. diff Chek-60 266 Stools of patients suspected of having CDIsubmitted to laboratories of two acute-carehospitals

Unclear 9.0

Turgeon [65] 2003 USA CCNA ImmunoCard C. difficile 1003 Consecutive stools of adults and childrensuspected of having CDI at five major hospitals

Unformed andformed

10.1

Vanpoucke [68] 2001 Belgium CCNA Ridascreen 156 Stool specimens submitted to laboratory ofuniversity hospital with request for CDI testing

Unformed 31.8

Viala [69] 2012 France TC BD GeneOhm, Xpert,Illumigene

94 Fresh stool specimens from symptomaticpatients collected at university hospital, 45 TCpositive and 49 TC negative were selected

Unformed 47.8

Walkty [70] 2013 Canada TC Illumigene, C. diff Quik Chek 428 All diarrhoeal stool specimens from patients>1 year old submitted for CDI testing to threemicrobiological laboratories serving majorhospitals and surrounding communities

Unformed 14.7

Wren [71] 2009 UK Culture, TC C. diff Quik Chek, Tox A/BQuik Chek

1007 Stool samples submitted for CDI testing frompatients who developed diarrhoea after beingadmitted to major university hospitals

Unformed 8.6

Zheng [72] 2004 USA Culture C. diff Chek-60 992 Stool samples submitted for routine CDI testingbecause of antibiotic-associated diarrhoeacollected from hospital laboratories andsupplied to TechLab, a large medical centre andreference laboratory

Unclear 13.8

CCNA, cell cytotoxicity neutralization assay; CDI, Clostridium difficile infection; TC, toxigenic culture.a Criteria were: oncology/haematology patient, specific request for CDI testing by treating physician, history of diarrhoea developed while receiving antibiotics, or pseudomembranous colitis.b Criteria were: all unformed faecal samples not clearly attributable to an underlying disease, or treatment from all hospital patients >2 years and from individuals in the community >65 years irrespective of C. difficile or other

testing requests.

M.J.T.Crobach

etal./


andInfection

xxx(2016)

1e19

10Pleasecite

thisarticle

inpress


SocietyofClinicalM

icrobiologyand


ofthediagnostic

guidancedocum

entfor



andInfection


0

10

20

30

40

50

60

70

80

90

100

Not described or partly described

No

Partly incorrect

Yes

Fig. 2. Quality assessment of included studies.


Toxin A/B EIAs tended to be the most specific assays, while GDHEIAs and NAATsweremore sensitive tests. Althoughmany toxin A/BEIAs belong to the least sensitive tests, the sensitivity of this cate-gory of assays is not as low as reported earlier [1]. This is becauseonly currently available tests were included in the present analysis,and the newer generation of toxin A/B EIAs turns out to be moresensitive than the earlier toxin A EIAs.

We compared all categories of the index tests (GDH EIAs, toxinA/B EIAs and NAATs) to both of the reference tests, CCNA and TC.However, not only are the targets of these three categories of indextests somewhat different, but also the targets of the two referencetests differ: CCNA detects in vivo toxin production, while TC detectsthe presence of a toxigenic C. difficile strain.

This explains why sensitivities and specificities were differentfor each reference test that was used as a comparator. For example,toxin A/B EIAs were less sensitive compared to TC instead of CCNA:toxin EIAs will not (like the TC) detect all samples containingtoxigenic C. difficile strains but only (some of) those with free toxinpresent. It also explains why NAATs were less specific compared toCCNA instead of TC: NAATs are not able (like CCNA) to discernsamples with in vivo toxin production from samples with in vitrotoxin production.

We included both CCNA and TC as reference tests, as there hasalways been debate which of these tests best defines CDI cases.Recently a large study reported that CCNA positivity (i.e. demon-stration of free toxin) but not TC positivity (i.e. demonstration oftoxin-producing capacity) correlated with clinical outcome.Therefore, at least all samples with a positive CCNA can beconsidered to represent true CDI cases [47]. However, samples witha positive TC but negative CCNA are difficult to interpret. Thesesamples could either belong to C. difficile carriers (harbouring atoxigenic C. difficile strain not producing detectable toxins at thatmoment) or to patients with CDI with toxin levels below thethreshold of detection.


To guarantee a certain level of uniformity and quality, onlystudies that met our eligibility criteria were included in the meta-analysis. Still, studies differed from one another in many aspects.For CCNA, diverse dilutions of faecal filtrate and diverse cell lineswere used. For TC, diverse culture media and diverse methods todemonstrate toxigenicity were applied. Also, none of the studiessatisfied all our quality assessment criteria. Notwithstanding thesedifferences, all included studies met the minimaldquitestrictdrequirements we set. We therefore think that it is justifiablethat we calculated summary estimates of sensitivity and specificity,especially because we intended to provide a general overview oftest performances of different categories of laboratory assaysinstead of pointing out one ‘best’ assay. It is, however, important torealize that test performances of individual assays may have beeninfluenced by the design of included studies analysing these tests.Besides, test characteristics presented here should not be consid-ered unchanging over time and should not be considered fixedcharacteristics. This is because procedures of commercial assays aresometimes revised to enhance test performance, and also becauseassays may perform differently among different populations (e.g.high- vs. low-risk patients). Also, in all categories, new assays weremarketed. The introduction of newer toxin A/B EIAs leading to abetter sensitivity of this category of assays is a good example of thelatter.

On the basis of the review results, PPVs and NPVs were calcu-lated at different hypothetical prevalences of CDI in the testedpopulation. The prevalence of CDI can be seen as the pretestprobability of having CDI and would typically be around 5e10% inan endemic setting [73]. At a CDI prevalence of 5%, even the mostspecific tests (toxin A/B EIAs) would have PPVs of only 69e81%. Onthe contrary, NPVs would be very high for all index tests. If theprevalence of CDI would rise to 50% among the tested patients, thePPV would consequently raise to 98.8% for the most specific test,but the NPV would drop to 82.5% for the least sensitive tests. Both


Table 5Sensitivity and specificity of index tests compared to CCNA

Type Index test Study Sensitivity (95% CI) Specificity (95% CI)

(A) Well-type EIA GDH C. diff Chek-60 Eastwood [27] 0.90 (0.82e0.95) 0.93 (0.90e0.95)C. diff Chek-60 Planche [47] 0.96 (0.95e0.98) 0.92 (0.92e0.93)C. diff Chek-60 Qutub [48] 0.94 (0.84e0.99) 0.88 (0.80e0.94)C. diff Chek-60 Reller [50] 0.91 (0.79e0.98) 0.90 (0.87e0.92)C. diff Chek-60 Ticehurst [64] 0.96 (0.79e1.00) 0.90 (0.86e0.94)

(B) Membrane-type EIA GDH C. diff Quik Chek Eckert [8] 1.00 (0.85e1.00) 0.92 (0.88e0.94)C. diff Quik Chek Larson [36] 1.00 (0.92e1.00) 0.90 (0.87e0.92)C. diff Quik Chek Reller [50] 1.00 (0.92e1.00) 0.83 (0.79e0.86)ImmunoCard C. difficile Staneck [60] 0.84 (0.77e0.90) 0.92 (0.90e0.94)ImmunoCard C. difficile Turgeon [65] 0.80 (0.71e0.87) 0.92 (0.91e0.94)Quik Chek Complete-GDH Ota [45] 0.81 (0.61e0.93) 0.82 (0.73e0.88)Quik Chek Complete-GDH Swindells [61] 1.00 (0.78e1.00) 0.95 (0.90e0.98)

(C) Well-type EIA toxins A/B Clostridium difficile Tox A/B II Eastwood [27] 0.91 (0.84e0.95) 0.96 (0.93e0.97)Clostridium difficile Tox A/B II Massey [39] 0.75 (0.67e0.82) 0.98 (0.96e0.99)Clostridium difficile Tox A/B II Musher [41] 0.96 (0.87e1.00) 0.87 (0.77e0.94)Clostridium difficile Tox A/B II O'Connor [44] 0.80 (0.68e0.89) 0.99 (0.96e1.00)Clostridium difficile Tox A/B II Planche [47] 0.83 (0.80e0.86) 0.99 (0.99e0.99)Clostridium difficile Tox A/B II Qutub [48] 0.73 (0.59e0.84) 1.00 (0.96e1.00)Premier toxins A/B Berg, van den 2007 [67] 0.97 (0.83e1.00) 0.94 (0.92e0.96)Premier toxins A/B Eastwood [27] 0.92 (0.85e0.96) 0.97 (0.95e0.98)Premier toxins A/B Musher [41] 0.99 (0.93e1.00) 0.97 (0.95e0.99)Premier toxins A/B O'Connor [44] 0.82 (0.70e0.91) 0.99 (0.96e1.00)Premier toxins A/B Ota [45] 0.58 (0.37e0.77) 1.00 (0.97e1.00)Premier toxins A/B Planche [47] 0.67 (0.63e0.71) 0.99 (0.99e0.99)Remel ProSpecT Eastwood [27] 0.90 (0.83e0.95) 0.93 (0.90e0.95)Remel ProSpecT Musher [41] 0.91 (0.80e0.97) 0.97 (0.91e1.00)Ridascreen toxins A/B Eastwood [27] 0.67 (0.57e0.75) 0.95 (0.93e0.97)Ridascreen toxins A/B Vanpoucke [68] 0.57 (0.43e0.70) 0.97 (0.92e0.99)

(D) Membrane-type EIA toxins A/B ImmunoCard toxins A/B Berg, van den (2005) [66] 0.91 (0.72e0.99) 0.97 (0.95e0.99)ImmunoCard toxins A/B Eastwood [27] 0.85 (0.76e0.91) 0.99 (0.98e1.00)ImmunoCard toxins A/B Musher [41] 0.96 (0.89e0.99) 0.99 (0.97e1.00)Quik Chek Complete-Tox A/B Ota [45] 0.50 (0.30e0.70) 1.00 (0.97e1.00)Quik Chek Complete-Tox A/B Swindells [61] 0.73 (0.45e0.92) 1.00 (0.97e1.00)Tox A/B Quik Chek Eastwood [27] 0.84 (0.76e0.91) 0.99 (0.98e1.00)Tox A/B Quik Chek Reller [50] 0.61 (0.45e0.75) 0.99 (0.98e1.00)Xpect Boer, de [25] 0.44 (0.20e0.70) 1.00 (0.97e1.00)Xpect Eastwood [27] 0.83 (0.74e0.90) 0.99 (0.98e1.00)

(E) Automated EIA toxins A/B VIDAS CDAB Eastwood [27] 0.98 (0.93e1.00) 0.99 (0.98e1.00)VIDAS CDAB Swindells [61] 0.53 (0.27e0.79) 1.00 (0.97e1.00)

(F) NAAT Amplivue Eckert [8] 0.96 (0.78e1.00) 0.95 (0.91e0.97)BD GeneOhm Eastwood [27] 0.92 (0.85e0.97) 0.95 (0.93e0.97)BD GeneOhm Stamper (2009e1) [59] 0.91 (0.78e0.97) 0.95 (0.92e0.97)BD GeneOhm Swindells [61] 1.00 (0.78e1.00) 0.98 (0.94e1.00)BD GeneOhm Terhes [63] 0.95 (0.82e0.99) 0.96 (0.94e0.98)Illumigene Noren [42] 1.00 (0.90e1.00) 0.93 (0.89e0.96)Illumigene Ota [45] 0.88 (0.70e0.98) 0.97 (0.93e0.99)Illumigene Pancholi [46] 0.87 (0.66e0.97) 0.91 (0.86e0.95)Prodesse ProGastro Cd assay Stamper (2009e2) [58] 0.83 (0.65e0.94) 0.96 (0.92e0.98)Xpert C. difficile Berry [20] 1.00 (0.94e1.00) 0.94 (0.92e0.95)Xpert C. difficile Huang [31] 0.96 (0.78e1.00) 0.87 (0.82e0.92)Xpert C. difficile Pancholi [46] 1.00 (0.85e1.00) 0.89 (0.83e0.93)Xpert C. difficile Planche [47] 0.98 (0.96e0.99) 0.93 (0.92e0.94)Xpert C. difficile Swindells [61] 1.00 (0.78e1.00) 0.97 (0.93e0.99)

CI, confidence interval; CCNA, cell cytotoxicity neutralization assay; EIA, enzyme immunoassay; GDH, glutamate dehydrogenase; NAAT, nucleic acid amplification test.


suboptimal PPV and NPV have implications. A low PPVwill result inmany patients with false-positive results. These noninfected pa-tients may receive unnecessary treatment for CDI, and unnecessaryisolation precautions may be taken. A low NPV will result in manyundetected cases, which may not only have implications for indi-vidual patients but also for further transmission of C. difficile. It istherefore important to be aware not only of the sensitivity andspecificity of an assay but also of the CDI prevalence in the testedpopulation, as the predictive values and hence the clinical utility ofthe assays depend on them.

The easiest way to diagnose CDI would be to use a single rapidlaboratory test that is able to reliably predict disease status. A rapidCDI diagnosis is associated with more prompt CDI treatment andless unnecessarily treated patients [74]. However, two problemsarise if the rapid assays are used as stand-alone test for diagnosing


CDI. First, as described above, the PPVs of even the most specifictests are inadequate at low disease prevalence. If toxin EIAs were tobe used in an endemic situation (CDI prevalence of 5% in the testedpopulation, PPV 81%), an unacceptably high percentage (19%) ofpatients with a positive test result would not actually have CDI.Second, as the targets identified by the index tests are (just like thetargets of the reference test) different from each other, a positiveindex test does not necessarily indicate a real CDI case. Two of thethree categories of index test are not able to differentiate carriersfrom CDI patients: both GDH EIAs and NAATs do not detect freetoxins. Using NAAT as a stand-alone test and relying on clinicalsymptoms to discern patients with CDI from asymptomatic carriersis not an optimal approach: patients colonized by a toxigenicC. difficile strain may very well develop diarrhoea due to othercauses, and no specific clinical symptoms exist to differentiate CDI


Table 6Sensitivity and specificity of index tests compared to TC


(A) Well-type EIA GDH C. diff Chek-60 Planche [47] 0.94 (0.93e0.96) 0.94 (0.94e0.95)(B) Membrane-type EIA GDH C. diff Quik Chek Eckert [8] 0.97 (0.85e1.00) 0.95 (0.92e0.97)

C. diff Quik Chek Walkty [70] 0.83 (0.71e0.91) 0.97 (0.95e0.98)ImmunoCard C. difficile Barkin [19] 1.00 (0.90e1.00) 1.00 (0.98e1.00)ImmunoCard C. difficile Jacobs [32] 0.60 (0.32e0.84) 0.76 (0.68e0.83)Quik Chek CompletedGDH Bruins [21] 0.97 (0.90e1.00) 0.98 (0.96e0.98)Quik Chek CompletedGDH Kawada [33] 1.00 (0.88e1.00) 0.88 (0.71e0.96)Quik Chek CompletedGDH Swindells [61] 1.00 (0.81e1.00) 0.97 (0.92e0.99)

(C) Well-type EIA toxins A/B Clostridium difficile Tox A/B II Planche [47] 0.58 (0.55e0.61) 0.99 (0.98e0.99)Clostridium difficile Tox A/B II Snell [56] 0.85 (0.72e0.93) 0.98 (0.96e0.99)Premier toxins A/B Barkin [19] 0.86 (0.71e0.95) 0.91 (0.86e0.94)Premier toxins A/B Bruins [21] 0.41 (0.30e0.53) 0.99 (0.98e0.99)Premier toxins A/B Leitner [38] 0.40 (0.21e0.61) 1.00 (0.98e1.00)Premier toxins A/B Novak-Weekley [43] 0.58 (0.46e0.70) 0.95 (0.92e0.97)Premier toxins A/B Planche [47] 0.46 (0.42e0.49) 0.99 (0.99e0.99)Premier toxins A/B Sloan [55] 0.48 (0.32e0.63) 0.98 (0.94e1.00)Ridascreen toxins A/B Mattner [40] 0.52 (0.36e0.68) 0.98 (0.95e0.99)

(D) Membrane-type EIA toxins A/B ImmunoCard toxins A/B Bruins [21] 0.41 (0.30e0.53) 0.99 (0.98e1.00)ImmunoCard toxins A/B de Jong [26] 0.47 (0.23e0.72) 0.99 (0.96e1.00)ImmunoCard toxins A/B Sloan [55] 0.48 (0.32e0.63) 0.99 (0.95e1.00)Quik Chek CompletedTox A/B Bruins [21] 0.55 (0.43e0.66) 1.00 (1.00e1.00)Quik Chek CompletedTox A/B Calderaro [23] 0.68 (0.55e0.80) 0.89 (0.84e0.92)Quik Chek CompletedTox A/B Hart [29] 0.29 (0.16e0.44) 1.00 (0.97e1.00)Quik Chek CompletedTox A/B Kawada [33] 0.79 (0.59e0.92) 0.97 (0.84e1.00)Quik Chek CompletedTox A/B Kim (2014) [35] 0.64 (0.50e0.76) 0.98 (0.96e0.99)Quik Chek CompletedTox A/B Swindells [61] 0.61 (0.36e0.83) 1.00 (0.97e1.00)Tox A/B Quik Chek Kawada [33] 0.71 (0.51e0.87) 0.94 (0.79e0.99)Tox A/B Quik Chek Le Guern [37] 0.43 (0.28e0.59) 1.00 (0.98e1.00)Tox A/B Quik Chek Wren [71] 0.40 (0.30e0.51) 1.00 (1.00e1.00)Xpect Sloan [55] 0.48 (0.32e0.63) 0.84 (0.77e0.89)

(E) Automated EIA toxins A/B VIDAS CDAB Jong, de [26] 0.71 (0.42e0.92) 0.95 (0.90e0.98)VIDAS CDAB Kim (2012) [34] 0.64 (0.31e0.89) 1.00 (0.97e1.00)VIDAS CDAB Kim (2014) [35] 0.76 (0.61e0.87) 0.97 (0.96e0.99)VIDAS CDAB Shin (2009e1) [52] 0.68 (0.62e0.73) 0.96 (0.95e0.97)VIDAS CDAB Shin (2009e2) [51] 0.69 (0.59e0.78) 0.97 (0.94e0.98)VIDAS CDAB Shin (2012e2) [54] 0.44 (0.30e0.60) 1.00 (0.98e1.00)VIDAS CDAB Swindells [61] 0.44 (0.22e0.69) 1.00 (0.97e1.00)

(F) NAAT Advansure CD Kim (2012) [34] 1.00 (0.72e1.00) 0.98 (0.94e1.00)Advansure CD Soh [57] 0.85 (0.65e0.96) 0.98 (0.95e1.00)Amplivue Eckert [8] 0.86 (0.71e0.95) 0.98 (0.95e0.99)BD GeneOhm Buchan [22] 0.97 (0.86e1.00) 0.98 (0.95e1.00)BD GeneOhm Hart [29] 0.89 (0.76e0.96) 0.99 (0.95e1.00)BD GeneOhm Le Guern [37] 0.95 (0.85e0.99) 1.00 (0.98e1.00)BD GeneOhm Shin (2012e1) [53] 0.96 (0.88e0.99) 0.97 (0.93e0.99)BD GeneOhm Swindells [61] 0.94 (0.73e1.00) 0.99 (0.96e1.00)BD GeneOhm Viala [69] 0.96 (0.85e0.99) 0.98 (0.89e1.00)BD Max Cdiff Le Guern [37] 0.98 (0.88e1.00) 1.00 (0.98e1.00)BD Max Cdiff Leitner [38] 0.96 (0.80e1.00) 0.99 (0.96e1.00)GenomEra Hirvonen [30] 1.00 (0.95e1.00) 0.99 (0.96e1.00)Illumigene Barkin [19] 1.00 (0.90e1.00) 1.00 (0.98e1.00)Illumigene Bruins [21] 0.93 (0.85e0.98) 1.00 (0.99e1.00)Illumigene Buchan [22] 0.93 (0.68e1.00) 0.95 (0.88e0.99)Illumigene Calderaro [23] 1.00 (0.94e1.00) 0.83 (0.78e0.87)Illumigene Hart [29] 0.89 (0.76e0.96) 1.00 (0.97e1.00)Illumigene Lalande [7] 0.92 (0.80e0.98) 0.99 (0.98e1.00)Illumigene Soh [57] 0.92 (0.75e0.99) 0.99 (0.97e1.00)Illumigene Viala [69] 0.87 (0.73e0.95) 1.00 (0.93e1.00)Illumigene Walkty [70] 0.73 (0.60e0.83) 1.00 (0.98e1.00)Portrait Buchan [22] 0.98 (0.94e1.00) 0.93 (0.90e0.95)Prodesse ProGastro Cd assay Stamper (2009e2) [58] 0.77 (0.62e0.89) 0.99 (0.97e1.00)Seeplex ACE Shin (2012e1) [53] 0.90 (0.80e0.96) 0.97 (0.93e0.99)Verigene Caroll [24] 0.91 (0.87e0.94) 0.93 (0.91e0.94)Xpert C. difficile Buchan [22] 1.00 (0.94e1.00) 0.92 (0.87e0.95)Xpert C. difficile Novak-Weekley [43] 0.94 (0.86e0.98) 0.96 (0.94e0.98)Xpert C. difficile Planche [47] 0.95 (0.93e0.96) 0.96 (0.96e0.97)Xpert C. difficile Shin (2012e2) [54] 1.00 (0.93e1.00) 0.95 (0.91e0.98)Xpert C. difficile Swindells [61] 1.00 (0.81e1.00) 0.99 (0.96e1.00)Xpert C. difficile Tenover [62] 0.93 (0.90e0.96) 0.94 (0.93e0.95)Xpert C. difficile Viala [69] 0.98 (0.88e1.00) 0.98 (0.89e1.00)

CI, confidence interval; EIA, enzyme immunoassay; GDH, glutamate dehydrogenase; NAAT, nucleic acid amplification test; TC, toxigenic culture.


Please cite this article in press as: CrobachMJT, et al., European Society of Clinical Microbiology and Infectious Diseases: update of the diagnosticguidance document for Clostridium difficile infection, Clinical Microbiology and Infection (2016), http://dx.doi.org/10.1016/j.cmi.2016.03.010

Table 7Sensitivity and specificity of index tests compared to culture


(A) Well-type EIA GDH C. diff Chek-60 Fenner [28] 0.93 (0.88e0.97) 0.97 (0.95e0.97)C. diff Chek-60 Reller (2007) [49] 1.00 (0.98e1.00) 0.67 (0.61e0.72)C. diff Chek-60 Snell [56] 0.94 (0.86e0.98) 0.98 (0.96e0.99)C. diff Chek-60 Zheng [72] 0.71 (0.63e0.78) 0.88 (0.85e0.90)

(B) Membrane-type EIA GDH C. diff Quik Chek Wren [71] 0.95 (0.90e0.98) 0.99 (0.98e1.00)Quik Chek CompletedGDH Bruins [21] 0.95 (0.89e0.99) 0.99 (0.98e0.99)Quik Chek CompletedGDH Hart [29] 0.87 (0.75e0.95) 0.97 (0.91e0.99)Quik Chek CompletedGDH Kawada [33] 1.00 (0.88e1.00) 0.93 (0.78e0.99)Quik Chek CompletedGDH Swindells [61] 1.00 (0.82e1.00) 0.98 (0.93e1.00)ImmunoCard C. difficile Jacobs [32] 0.75 (0.59e0.87) 0.90 (0.83e0.95)ImmunoCard C. difficile Kawada [33] 0.80 (0.61e0.92) 1.00 (0.88e1.00)

CI, confidence interval; EIA, enzyme immunoassay; GDH, glutamate dehydrogenase; NAAT, nucleic acid amplification test.

Table 8Pooled sensitivities and specificities of categories of tests

Type Test Compared to CCNA Compared to TC Compared to culture

No. ofstudies

Sensitivity(95% CI)

Specificity(95% CI)

No. ofstudies

Sensitivity(95% CI)

Specificity(95% CI)

No. ofstudies

Sensitivity(95% CI)

Specificity(95% CI)

EIA GDH Total 12 0.94 (0.89e0.97) 0.90 (0.88e0.92) 8 0.96 (0.86e0.99) 0.96 (0.91e0.98) 11 0.94 (0.86e0.97) 0.96 (0.92e0.98)Well type 5 0.94 (0.91e0.97) 0.92 (0.92e0.93) 1 0.94 (0.93e0.96) 0.94 (0.94e0.95) 4 0.89 (0.86e0.91) 0.91 (0.90e0.92)Membranetype

7 0.98 (0.78e1.00) 0.90 (0.87e0.93) 7 0.97 (0.84e1.00) 0.96 (0.90e0.99) 7 0.93 (0.84e0.97) 0.98 (0.95e0.99)

EIAtoxins A/B

Total 27 0.83 (0.76e0.88) 0.99 (0.98e0.99) 29 0.57 (0.51e0.63) 0.99 (0.98e0.99)Well type 18 0.85 (0.77e0.91) 0.98 (0.96e0.99) 16 0.60 (0.52e0.68) 0.98 (0.97e0.99)Membranetype

9 0.79 (0.66e0.88) 0.99 (0.98e0.99) 13 0.53 (0.45e0.61) 0.99 (0.97e1.00)

NAAT 14 0.96 (0.93e0.98) 0.94 (0.93e0.95) 32 0.95 (0.92e0.97) 0.98 (0.97e0.99)

CI, confidence interval; CCNA, cell cytotoxicity neutralization assay; EIA, enzyme immunoassay; GDH, glutamate dehydrogenase; NAAT, nucleic acid amplification test; TC,toxigenic culture.

Table 9PPV and NPV for different categories of index tests at hypothetical CDI prevalences of 5, 10, 20 and 50%

Test type CDI prevalence 5% CDI prevalence 10% CDI prevalence 20% CDI prevalence 50%

PPV NPV PPV NPV PPV NPV PPV NPV

Well-type EIA GDH 38 100 54 99 72 98 91 94Membrane-type EIA GDH 34 100 52 100 71 99 91 98Well-type EIA toxins A/B 69 99 83 98 91 96 98 87Membrane-type EIA toxins A/B 81 99 90 98 95 95 99 83NAAT 46 100 64 100 80 99 94 96

Pooled estimates of sensitivity and specificity compared to cell cytotoxicity neutralization assay were used to calculate the predictive values.CDI, Clostridium difficile infection; EIA, enzyme immunoassay; GDH, glutamate dehydrogenase; NAAT, nucleic acid amplification test; NPV, negative predictive value; PPV,positive predictive value.


from other causes of diarrhoea. From the above, we conclude thatneither GDH EIA nor toxin A/B EIA or NAAT can reliably be used as astand-alone test to diagnose CDI.

Because no single test is suitable to be used as a stand-alone test,it is best to combine two tests in an algorithm in order to optimizethe diagnosis of CDI. The advantage of an algorithm is that tests canbe combined in such a way that the percentage of false-positiveresults can be decreased. This can be done by testing all sampleswith a first test, then performing reflex testing on samples with apositive first test result only. The first test should be a test thatreliably classifies samples with a negative test result as non-CDI;these samples will not be tested further. This first test shouldtherefore be a test with a high NPV (i.e. a highly sensitive test).Thus, in our case, this first test can either be a GDH EIA or NAAT. Thechoice between these two categories of assays can be made by eachindividual laboratory. The second test should be a test with a highPPV (i.e. a highly specific test), so that all samples with a positivesecond test result can reliably be classified as CDI. Toxin A/B EIAscan very well be used for this purpose, because besides being the


most specific tests, these tests also have the advantage of detectingfree toxin. Thus, after application of a first sensitive test (GDH EIA orNAAT), the toxin A/B EIA can then be performed as a second step onall samples that tested positive by NAAT or GDH EIA (Fig. 3(a)).Samples with a positive second test result can be classified as CDIlikely to be present. However, samples with a first positive testresult but a negative toxin A/B EIA need to be clinically evaluated.Among these samples, CDI (with toxin levels below the threshold ofdetection or a false-negative toxin A/B EIA result) or C. difficilecarriage is possible.

A recent large study tried to establish the optimum diagnosticalgorithm for CDI [47]. In this study, 12 420 faecal samples weretested by diverse commercial assays, TC and CCNA. The overallperformance of combined tests was superior to individual tests. Thecombination of a NAAT (Xpert) and toxin A/B EIA (Techlab Tox A/BII) was the optimal algorithm compared to the CCNA test, but theGDH EIA (C. diff Chek-60)etoxin A/B EIA algorithm performedalmost identically [47]. These findings can be seen as a validation ofour more theoretical approach to establish the best testing strategy,


Step 1:Highly sensitive test: NAAT or GDH EIA

Step 2:Highly specific test:

Toxin A/B EIA

No further testing required:CDI is unlikely to be

present

CDI is likely to be present

Clinical evaluation: CDI or carriage of (toxigenic) C.

difficile is possible

Negative test resultPositive test result

Positive test result Negative test result

Step 3 (optional):Perform TC or NAAT (in

case first test was a GDH EIA)

Step 1:Highly sensitive test: GDH and Tox A/B

EIA

No further testing required: CDI is likely to

be present

CDI is unlikely to be present

Clinical evaluation: CDI or carriage of (toxigenic) C.

difficile is possible

Both positiveBoth negative

Negative test result

No further testing required: CDI is

unlikely to be present

GDH positive, ToxA/B negative

Step 2 (optional):NAAT or TC

Positive test result

(a)

(b)

Fig. 3. Recommended algorithms for CDI testing. (a) GDH or NAATeTox A/B algorithm. (b) GDH and Tox A/BeNAAT/TC algorithm. CDI, Clostridium difficile infection; GDH, glutamatedehydrogenase; NAAT, nucleic acid amplification test; TC, toxigenic culture; Tox A/B, toxin A/B; EIA, enzyme immunoassay.


and they endorse the conclusion that NAATetoxin A/B EIA, oralternatively GDH EIAetoxin A/B EIA, are two of the best algorithmsto diagnose CDI (Fig. 3(a)).

An alternative algorithm is to test simultaneously with both aGDH and toxin A/B EIA. An assay is available that includes boththese targets in one system (C. diff Quik Chek Complete; Techlab),but the sensitivity of the toxin component is unclear and may notbe as a high as some individual toxin EIAs (Tables 5e7). Samplesthat test negative for both GDH and toxin A/B can reliably beclassified as non-CDI, while samples that test positive for both GDH


and toxin A/B can be classified as CDI likely to be present. Sampleswith a GDH-positive result but that are negative for toxin couldundergo reflex testing by NAAT to determine if a toxigenic C. difficilestrain is present (Fig. 3(b)). Samples with a negative GDH result butthat are positive for toxin need to be retested, as this is an invalidresult. Only one study evaluating this kind of algorithm andcomparing it to a reference test was identified in the literature [45].In this specific study, samples were screened by C. diff Quik ChekComplete, and inconclusive results underwent reflex testing byIllumigene. The overall sensitivity for this algorithm compared to



CCNA was 81%, while specificity was reported to be 100%. Theoverall sensitivity and specificity of this and the aforementionedalgorithm depend, however, on the individual assays that areincluded.

Although we recommend the use of an algorithm for CDI testingbased on two rapid assays, every laboratory should also be able toisolate C. difficile, ideally via TC from selected samples, for tworeasons. First, TC offers the ability to perform molecular typing andsusceptibility testing on recovered isolates from positive samplesand can be used for outbreak investigations [75]. Second, sampleswith a positive GDH EIA and/or NAAT but a negative toxin A/B EIAmay either be samples that tested falsely positive on GDH EIA/NAATor samples containing C. difficile, but without detectable free toxin.To be able to discern between these two conditions, a third-stagereflex test to either a TC or NAAT or GDH (if not yet performed)can be performed on samples with discordant results. For patientswith evidence of C. difficile but negative toxin A/B EIA, clinicalevaluation is needed, and clinical considerations come into play todetermine a case as either positive or negative; these patients caneither be CDI patients with undetectable toxin levels, or false-negative toxin A/B EIA results or potential carriers of toxigenicC. difficile. Although C. difficile carriers may play an important role inthe spread of the disease [76,77], the indication for treating thesepatients for CDI remains controversial. In addition, the need forisolation precautions for these patients remains to be clarified.Therefore, performing TCs on these samples can be of importancefor epidemiologic purposes, but it is not yet a prerequisite for pa-tient management.

The decision to treat CDI is ultimately a clinical decision, guidedby laboratory results. No tests are infallible, so it may be clinicallyjustified to treat a patient for CDI despite negative test results;treatment should not be withheld on the basis of laboratory testsalone. However, patients with toxin-negative specimens shouldhave alternative diagnoses considered and excluded; provided anadequate testing strategy is followed, most patients with negativeresults for CDI will truly not have this infection, and thus treatmentwill be unnecessary.

Besides the question which assay or algorithm should be usedfor CDI detection, another issue is the number of specimens perpatient that should be submitted for testing. Before the introduc-tion of algorithms to diagnose CDI, lack of confidence in the tests forCDI detection (mainly toxin EIAs) led to the practice of multiplesample submission. However, the diagnostic gain of repeat testingwithin a 7-day period with both toxin A/B EIA and PCR wasdemonstrated to be very low [78]. If one of the above proposedalgorithms is used, then the adequate NPV at low disease preva-lence is based on original studies which did not test samplesrepeatedly by index test and only once by reference test. Thisadequate NPV indicates that routine submission of multiple sam-ples after a first negative test round has to be discouraged; thesesamples can reliably be classified as non-CDI.

However, in cases of ongoing clinical suspicion during anendemic situation, the submission of a repeat sample may bejustified, as these specific algorithms will have adequate PPVs evenin a low-prevalence situation.

In outbreak situations with a higher CDI prevalence in the testedpopulation, the NPV of the algorithm will fall. In such an outbreaksituation, submitting a repeat sample in case of ongoing clinicalsuspicionwill be of value, as has been shown for toxin A/B EIA [79].

Testing for cure is not recommended, as patients can shedspores and even toxins of C. difficile for a prolonged time afterresolution of diarrhoea [80,81]. The infection can be consideredresolved when symptoms of diarrhoea have resolved.

Selection of which of submitted stool samples should be testedfor CDI is also important. Recognition of potential CDI cases may be


burdensome, as it is increasingly being recognized that CDI is notonly acquired in healthcare facilities by patients with well-knownrisk factors for the disease. In the Netherlands, C. difficile wasrelatively frequent among patients with diarrhoeal complaints ingeneral practice [82]. Community-onset CDI can affect all agegroups, and many patients do not have known risk factors [83,84].A recent study showed that on a single day in Spain, two of everythree CDI episodes were underdiagnosed ormisdiagnosed owing tononsensitive tests (19.%) but more importantly to lack of suspicionand request (47.6%) [85]. Especially for nonhospitalized patientsand younger patients, CDI tests were not requested [85]. This trendwas also seen in a study involving almost 500 hospitals in 20countries across Europe: on two sampling days, 23% of sampleswith a positive CDI test result were initially missed due to lack ofsuspicion [73]. Hence, restricting testing to samples with a physi-cian's request for CDI testing will lead to underdiagnosis.

Empirical testing of all unformed stool samples submitted to thelaboratory was shown to increase the diagnostic yield [73,86]. Werecommend testing all unformed faecal samples submitted to thelaboratory (except samples from children under age 3). In infants,high rates of asymptomatic colonization with both toxigenic andnontoxigenic strains have been described [87]. Even in the case oftoxin production, infants rarely develop clinical disease. However,CDI can occur in infants and young children [88]. A recentlyreleased policy statement from the American Academy of Pediatricsrecommends to test for CDI only if age-specific clinical criteria aremet [14]. According to their statement, searching for alternativeaetiologies should be performed even in the case of a positive CDItest for children under 3 years of age. Concerning the problematicinterpretation of positive test results in this population, we indeedrecommend to limit testing of samples from children under age 3 tosamples with a physician's request only. Unformed stool samples ofchildren 3 years and older can be managed in the same way asdescribed above.

Clinical signs and symptoms are essential to CDI diagnosis.Therefore, formed stool samples should not be tested for CDI, asthese do not meet the clinical criteria of CDI. However, sometimesonly solid parts of diarrhoeal faecesmay be collected and submittedfor C. difficile testing. Local protocols therefore need to enableC. difficile testing on specific samples to take place. Also, anexception has to be made for patients suspected of CDI who haveileus. In these patients, a rectal swab can be used with adequatesensitivity and specificity for (toxigenic) culture, NAAT or GDH EIA[89,90]. The use of perirectal swabs for NAAT or GDH EIA testingmight also be an alternative in selected patient populations butmay depend on the presence of faecal staining of the swab [89e91].However, the use of (peri)rectal swabs has not been evaluated fortoxin EIA, and therefore clinical judgement remains essential inthese cases to discern colonized patients from patients with CDI.

Recommendations

Sample selection

� We recommend that CDI testing should not be limited to sam-ples with a specific physician's request. (Strong recommenda-tion, high-quality evidence)

� We suggest that at least all submitted unformed stool samplesfrom patients 3 years or older should be tested for CDI. (Weakrecommendation, low-quality evidence)

� We suggest to limit testing of samples from children under age 3to samples with a physician's request only. (Weak recommen-dation, low-quality evidence)

� Formed stool samples should not be tested for CDI (except incase of paralytic ileus). (Good practice statement)



� In patients suspected of ileus, a rectal swab can be used for(toxigenic) culture, NAAT or GDH EIA. (Strong recommendation,moderate-quality evidence)

Testing protocol

� The diagnosis of CDI should be based on clinical signs andsymptoms in combination with laboratory tests. Decision fortreatment for CDI is a clinical decision and may be justified evenif all laboratory tests are negative. (Good practice statement)

� We recommend against the use of a single rapid test as a stand-alone test due to inadequate PPV in an endemic situation.(Strong recommendation, moderate-quality evidence)

� We recommend the use of a 2-step algorithm (Fig. 3(A)). (Strongrecommendation, moderate-quality evidence)

� This algorithm should start with either NAAT or GDH EIA.Samples with a negative first test result can be reported asnegative. (Strong recommendation, moderate-quality evidence)

� Samples with a positive first test result should be tested furtherwith a toxin A/B EIA. Samples with a positive second test resultscan be reported as CDI-positive. (Strong recommendation,moderate-quality evidence)

� An alternative algorithm is to screen samples with both a GDHand toxin A/B EIA (Fig. 3(B)). Samples with concordant positiveor negative results can be reported as such. Samples with anegative GDH result but positive for toxin need to be retested asthis is an invalid result. (Strong recommendation, moderate-quality evidence)

� Samples with a positive first test result and negative second testresult (Fig. 3(A)) and samples with a GDH-positive test result butnegative toxin A/B test result (Fig. 3(B)) may represent sampleswith CDI or C. difficile carriage andmay optionally be tested withTC or NAAT (if not performed yet). (Weak recommendation,moderate-quality evidence)

� We recommend to perform TC and molecular typing of recov-ered isolates in case of outbreak situations. (Good practicestatement)

Repeated testing

� Repeated testing after a first positive sample during the samediarrhoeal episode is not recommended in an endemic situation.(Strong recommendation, moderate-quality evidence)

� Repeated testing after a first negative sample during the samediarrhoeal episode may be useful in selected cases with ongoingclinical suspicion during an epidemic situation or in cases withhigh clinical suspicion during endemic situations. (Strongrecommendation, moderate-quality evidence)

� A test of cure is not recommended. (Good practice statement)

Acknowledgements

Members of the Executive Committee are as follows: M. J. T.Crobach, Department of Medical Microbiology, Centre for InfectiousDiseases, Leiden University Medical Center, Leiden, TheNetherlands; T. Planche, Department of Medical Microbiology, StGeorge's Hospital, London, UK; C. Eckert and F. Barbut, ESGCDmembers, National Reference Laboratory for Clostridium difficile,Paris, France; O. M. Dekkers, Departments of Clinical Epidemiologyand Internal Medicine, Leiden University Medical Center, Leiden,The Netherlands, and Department of Clinical Epidemiology, AarhusUniversity, Aarhus, Denmark; E. M. Terveer, Department of MedicalMicrobiology, Centre for Infectious Diseases, Leiden UniversityMedical Center, Leiden, The Netherlands; M. H. Wilcox, ESGCDmember, Department of Microbiology, Leeds Teaching Hospitals &


University of Leeds, Leeds, UK; E. J. Kuijper (chair), ESGCD member,Department of Medical Microbiology, Centre for Infectious Dis-eases, Leiden University Medical Center, Leiden, The Netherlands.

Members of the Drafting Group are as follows: F. Allerberger,ESGCD member, Austrian Agency for Health and Food Safety(AGES), Vienna, Austria; L. von Müller, ESGCD member, Institut fürLabormedizin, Mikrobiologie und Hygiene (LMH), Christophorus-Kliniken GmbH, Coesfeld, and Advisory Laboratory for Clostridiumdifficile, Homburg/Saar, Germany; F. Fitzpatrick, ESGCD member,Department of Clinical Microbiology, Microbiology Laboratory,Beaumont Hospital, Ireland; R. Frei, ESGCD member, Division ofClinical Microbiology, University Hospital Basel, Basel, Switzerland;S. Mentula, ESGCD member, Bacterial Infections Unit, NationalInstitute for Health andWelfare, Helsinki, Finland; J. E. Coia, ESGCDmember, Department of Clinical Microbiology, Glasgow Royal In-firmary, Glasgow, UK; T. Nor�en, ESGCD member, Department ofInfection Control, €Orebro University Hospital, €Orebro, Sweden; J.van Broeck and M. Delmee, ESGCD members, Universit�e Catholiquede Louvain, Microbiology Unit, Brussels, Belgium.

Appendix A. Supplementary data

Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.cmi.2016.03.010.

Transparency Declaration

All authors report no conflicts of interest relevant to this article.

References

[1] Crobach MJ, Dekkers OM, Wilcox MH, Kuijper EJ. European Society of ClinicalMicrobiology and Infectious Diseases (ESCMID): data review and recom-mendations for diagnosing Clostridium difficileeinfection (CDI). Clin MicrobiolInfect 2009;15:1053e66.

[2] Planche T, Wilcox M. Reference assays for Clostridium difficile infection: one ortwo gold standards? J Clin Pathol 2011;64:1e5.

[3] Delmee M. Laboratory diagnosis of Clostridium difficile disease. Clin MicrobiolInfect 2001;7:411e6.

[4] Hink T, Burnham CA, Dubberke ER. A systematic evaluation of methods tooptimize culture-based recovery of Clostridium difficile from stool specimens.Anaerobe 2013;19:39e43.

[5] Borriello SP, Honour P. Simplified procedure for the routine isolation ofClostridium difficile from faeces. J Clin Pathol 1981;34:1124e7.

[6] Burnham CA, Carroll KC. Diagnosis of Clostridium difficile infection: an ongoingconundrum for clinicians and for clinical laboratories. Clin Microbiol Rev2013;26:604e30.

[7] Lalande V, Barrault L, Wadel S, Eckert C, Petit JC, Barbut F. Evaluation of a loop-mediated isothermal amplification assay for diagnosis of Clostridium difficileinfections. J Clin Microbiol 2011;49:2714e6.

[8] Eckert C, Holscher E, Petit A, Lalande V, Barbut F. Molecular test based onisothermal helicase-dependent amplification for detection of the Clostridiumdifficile toxin A gene. J Clin Microbiol 2014;52:2386e9.

[9] Nyc O, Pituch H, Matejkova J, Obuch-Woszczatynski P, Kuijper EJ. Clostridiumdifficile PCR ribotype 176 in the Czech Republic and Poland. Lancet2011;377(9775):1407.

[10] Whiting PF, Rutjes AW, Westwood ME, Mallett S, Deeks JJ, Reitsma JB, et al.QUADAS-2: a revised tool for the quality assessment of diagnostic accuracystudies. Ann Intern Med 2011;155:529e36.

[11] Cohen SH, Gerding DN, Johnson S, Kelly CP, Loo VG, McDonald LC, et al.Clinical practice guidelines for Clostridium difficile infection in adults: 2010update by the Society for Healthcare Epidemiology of America (SHEA) and theInfectious Diseases Society of America (IDSA). Infect Control Hosp Epidemiol2010;31:431e55.

[12] Cheng AC, Ferguson JK, Richards MJ, Robson JM, Gilbert GL, McGregor A, et al.Australasian Society for Infectious Diseases guidelines for the diagnosis andtreatment of Clostridium difficile infection. Med J Aust 2011;194:353e8.

[13] Surawicz CM, Brandt LJ, Binion DG, Ananthakrishnan AN, Curry SR,Gilligan PH, et al. Guidelines for diagnosis, treatment, and prevention ofClostridium difficile infections. Am J Gastroenterol 2013;108:478e98.

[14] Schutze GE, Willoughby RE. Clostridium difficile infection in infants and chil-dren. Pediatrics 2013;131:196e200.

[15] Department of Health. Updated guidance on the diagnosis and reporting ofClostridium difficile. London: NHS; 2012.




http://refhub.elsevier.com/S1198-743X(16)30025-8/sref1


























































[16] Brouwers MC, Kho ME, Browman GP, Burgers JS, Cluzeau F, Feder G, et al.AGREE II: advancing guideline development, reporting and evaluation inhealth care. J Clin Epidemiol 2010;63:1308e11.

[17] Schunemann HJ, Oxman AD, Brozek J, Glasziou P, Jaeschke R, Vist GE, et al.Grading quality of evidence and strength of recommendations for diagnostictests and strategies. BMJ 2008;336(7653):1106e10.

[18] Guyatt GH, Schunemann HJ, Djulbegovic B, Akl EA. Guideline panels shouldnot GRADE good practice statements. J Clin Epidemiol 2015;68:597e600.

[19] Barkin JA, Nandi N, Miller N, Grace A, Barkin JS, Sussman DA. Superiority of theDNA amplification assay for the diagnosis of C. difficile infection: a clinicalcomparison of fecal tests. Dig Dis Sci 2012;57:2592e9.

[20] Berry N, Sewell B, Jafri S, Puli C, Vagia S, Lewis AM, et al. Real-time polymerasechain reaction correlates well with clinical diagnosis of Clostridium difficileinfection. J Hosp Infect 2014;87:109e14.

[21] Bruins MJ, Verbeek E, Wallinga JA, Bruijnesteijn van Coppenraet LE,Kuijper EJ, Bloembergen P. Evaluation of three enzyme immunoassays and aloop-mediated isothermal amplification test for the laboratory diagnosis ofClostridium difficile infection. Eur J Clin Microbiol Infect Dis 2012;31:3035e9.

[22] Buchan BW, Mackey TL, Daly JA, Alger G, Denys GA, Peterson LR, et al.Multicenter clinical evaluation of the Portrait toxigenic C. difficile assay fordetection of toxigenic Clostridium difficile strains in clinical stool specimens.J Clin Microbiol 2012;50:3932e6.

[23] Calderaro A, Buttrini M, Martinelli M, Gorrini C, Montecchini S, Medici MC,et al. Comparative analysis of different methods to detect Clostridium difficileinfection. New Microbiol 2013;36:57e63.

[24] Carroll KC, Buchan BW, Tan S, Stamper PD, Riebe KM, Pancholi P, et al.Multicenter evaluation of the Verigene Clostridium difficile nucleic acid assay.J Clin Microbiol 2013;51:4120e5.

[25] de Boer RF, Wijma JJ, Schuurman T, Moedt J, Dijk-Alberts BG, Ott A, et al.Evaluation of a rapid molecular screening approach for the detection oftoxigenic Clostridium difficile in general and subsequent identification of thetcdC Delta117 mutation in human stools. J Microbiol Methods 2010;83:59e65.

[26] de Jong E, de Jong AS, Bartels CJ, van der Rijt-van den Biggelaar C,Melchers WJ, Sturm PD. Clinical and laboratory evaluation of a real-time PCRfor Clostridium difficile toxin A and B genes. Eur J Clin Microbiol Infect Dis2012;31:2219e25.

[27] Eastwood K, Else P, Charlett A, Wilcox M. Comparison of nine commerciallyavailable Clostridium difficile toxin detection assays, a real-time PCR assay forC. difficile tcdB, and a glutamate dehydrogenase detection assay to cytotoxintesting and cytotoxigenic culture methods. J Clin Microbiol 2009;47:3211e7.

[28] Fenner L, Widmer AF, Goy G, Rudin S, Frei R. Rapid and reliable diagnosticalgorithm for detection of Clostridium difficile. J Clin Microbiol 2008;46:328e30.

[29] Hart J, Putsathit P, Knight DR, Sammels L, Riley TV, Keil A. Clostridium difficileinfection diagnosis in a paediatric population: comparison of methodologies.Eur J Clin Microbiol Infect Dis 2014;33:1555e64.

[30] Hirvonen JJ, Mentula S, Kaukoranta SS. Evaluation of a new automated ho-mogeneous PCR assay, GenomEra C. difficile, for rapid detection of toxigenicClostridium difficile in fecal specimens. J Clin Microbiol 2013;51:2908e12.

[31] Huang H, Weintraub A, Fang H, Nord CE. Comparison of a commercialmultiplex real-time PCR to the cell cytotoxicity neutralization assay fordiagnosis of Clostridium difficile infections. J Clin Microbiol 2009;47:3729e31.

[32] Jacobs J, Rudensky B, Dresner J, Berman A, Sonnenblick M, van DY, et al.Comparison of four laboratory tests for diagnosis of Clostridium difficile-eassociated diarrhea. Eur J Clin Microbiol Infect Dis 1996;15:561e6.

[33] Kawada M, Annaka M, Kato H, Shibasaki S, Hikosaka K, Mizuno H, et al.Evaluation of a simultaneous detection kit for the glutamate dehydrogenaseantigen and toxin A/B in feces for diagnosis of Clostridium difficile infection.J Infect Chemother 2011;17:807e11.

[34] Kim H, Jeong SH, Kim M, Lee Y, Lee K. Detection of Clostridium difficile toxin A/B genes by multiplex real-time PCR for the diagnosis of C. difficile infection.J Med Microbiol 2012;61(Pt 2):274e7.

[35] Kim H, Kim WH, Kim M, Jeong SH, Lee K. Evaluation of a rapid membraneenzyme immunoassay for the simultaneous detection of glutamate dehy-drogenase and toxin for the diagnosis of Clostridium difficile infection. Ann LabMed 2014;34:235e9.

[36] Larson AM, Fung AM, Fang FC. Evaluation of tcdB real-time PCR in a three-stepdiagnostic algorithm for detection of toxigenic Clostridium difficile. J ClinMicrobiol 2010;48:124e30.

[37] Le Guern, Herwegh S, Grandbastien B, Courcol R, Wallet F. Evaluation of a newmolecular test, the BD Max Cdiff, for detection of toxigenic Clostridium difficilein fecal samples. J Clin Microbiol 2012;50:3089e90.

[38] Leitner E, Einetter M, Grisold AJ, Marth E, Feierl G. Evaluation of the BD MAXCdiff assay for the detection of the toxin B gene of Clostridium difficile out offaecal specimens. Diagn Microbiol Infect Dis 2013;76:390e1.

[39] Massey V, Gregson DB, Chagla AH, Storey M, John MA, Hussain Z. Clinicalusefulness of components of the Triage immunoassay, enzyme immunoassayfor toxins A and B, and cytotoxin B tissue culture assay for the diagnosis ofClostridium difficile diarrhea. Am J Clin Pathol 2003;119:45e9.

[40] Mattner F, Winterfeld I, Mattner L. Diagnosing toxigenic Clostridium difficile:new confidence bounds show culturing increases sensitivity of the toxin A/Benzyme immunoassay and refute gold standards. Scand J Infect Dis 2012;44:578e85.


[41] Musher DM, Manhas A, Jain P, Nuila F, Waqar A, Logan N, et al. Detectionof Clostridium difficile toxin: comparison of enzyme immunoassay resultswith results obtained by cytotoxicity assay. J Clin Microbiol 2007;45:2737e9.

[42] Noren T, Alriksson I, Andersson J, Akerlund T, Unemo M. Rapid and sensitiveloop-mediated isothermal amplification test for Clostridium difficile detectionchallenges cytotoxin B cell test and culture as gold standard. J Clin Microbiol2011;49:710e1.

[43] Novak-Weekley SM, Marlowe EM, Miller JM, Cumpio J, Nomura JH, Vance PH,et al. Clostridium difficile testing in the clinical laboratory by use of multipletesting algorithms. J Clin Microbiol 2010;48:889e93.

[44] O'Connor D, Hynes P, Cormican M, Collins E, Corbett-Feeney G, Cassidy M.Evaluation of methods for detection of toxins in specimens of feces submittedfor diagnosis of Clostridium difficileeassociated diarrhea. J Clin Microbiol2001;39:2846e9.

[45] Ota KV, McGowan KL. Clostridium difficile testing algorithms using glutamatedehydrogenase antigen and C. difficile toxin enzyme immunoassays withC. difficile nucleic acid amplification testing increase diagnostic yield in atertiary pediatric population. J Clin Microbiol 2012;50:1185e8.

[46] Pancholi P, Kelly C, Raczkowski M, Balada-Llasat JM. Detection of toxigenicClostridium difficile: comparison of the cell culture neutralization, XpertC. difficile, Xpert C. difficile/Epi, and Illumigene C. difficile assays. J ClinMicrobiol 2012;50:1331e5.

[47] Planche TD, Davies KA, Coen PG, Finney JM, Monahan IM, Morris KA, et al.Differences in outcome according to Clostridium difficile testing method: aprospective multicentre diagnostic validation study of C. difficile infection.Lancet Infect Dis 2013;13:936e45.

[48] Qutub MO, AlBaz N, Hawken P, Anoos A. Comparison between the two-stepand the three-step algorithms for the detection of toxigenic Clostridiumdifficile. Indian J Med Microbiol 2011;29:293e6.

[49] Reller ME, Lema CA, Perl TM, Cai M, Ross TL, Speck KA, et al. Yield of stoolculture with isolate toxin testing versus a two-step algorithm including stooltoxin testing for detection of toxigenic Clostridium difficile. J Clin Microbiol2007;45:3601e5.

[50] Reller ME, Alcabasa RC, Lema CA, Carroll KC. Comparison of two rapid assaysfor Clostridium difficile Common antigen and a C difficile toxin A/B assay withthe cell culture neutralization assay. Am J Clin Pathol 2010;133:107e9.

[51] Shin BM, Lee EJ, Kuak EY, Yoo SJ. Comparison of VIDAS CDAB and CDAimmunoassay for the detection of Clostridium difficile in a tcdA� tcdBþ

C. difficile prevalent area. Anaerobe 2009;15:266e9.[52] Shin BM, Kuak EY, Lee EJ, Songer JG. Algorithm combining toxin immunoassay

and stool culture for diagnosis of Clostridium difficile infection. J Clin Microbiol2009;47:2952e6.

[53] Shin BM, Mun SJ, Yoo SJ, Kuak EY. Comparison of BD GeneOhm Cdiff andSeegene Seeplex ACE PCR assays using toxigenic Clostridium difficile culturefor direct detection of tcdB from stool specimens. J Clin Microbiol 2012;50:3765e7.

[54] Shin S, Kim M, Kim M, Lim H, Kim H, Lee K, et al. Evaluation of the XpertClostridium difficile assay for the diagnosis of Clostridium difficile infection. AnnLab Med 2012;32:355e8.

[55] Sloan LM, Duresko BJ, Gustafson DR, Rosenblatt JE. Comparison of real-timePCR for detection of the tcdC gene with four toxin immunoassays and cul-ture in diagnosis of Clostridium difficile infection. J Clin Microbiol 2008;46:1996e2001.

[56] Snell H, Ramos M, Longo S, John M, Hussain Z. Performance of the TechLab C.DIFF CHEK-60 enzyme immunoassay (EIA) in combination with the C. difficileTox A/B II EIA kit, the Triage C. difficile panel immunoassay, and a cytotoxinassay for diagnosis of Clostridium difficileeassociated diarrhea. J Clin Microbiol2004;42:4863e5.

[57] Soh YS, Yang JJ, You E, La JY, Kim MJ, Nam YS, et al. Comparison of two mo-lecular methods for detecting toxigenic Clostridium difficile. Ann Clin Lab Sci2014;44:27e31.

[58] Stamper PD, Babiker W, Alcabasa R, Aird D, Wehrlin J, Ikpeama I, et al. Eval-uation of a new commercial TaqMan PCR assay for direct detection of theClostridium difficile toxin B gene in clinical stool specimens. J Clin Microbiol2009;47:3846e50.

[59] Stamper PD, Alcabasa R, Aird D, Babiker W, Wehrlin J, Ikpeama I, et al.Comparison of a commercial real-time PCR assay for tcdB detection to a cellculture cytotoxicity assay and toxigenic culture for direct detection of toxin-producing Clostridium difficile in clinical samples. J Clin Microbiol 2009;47:373e8.

[60] Staneck JL, Weckbach LS, Allen SD, Siders JA, Gilligan PH, Coppitt G, et al.Multicenter evaluation of four methods for Clostridium difficile detection:ImmunoCard C. difficile, cytotoxin assay, culture, and latex agglutination. J ClinMicrobiol 1996;34:2718e21.

[61] Swindells J, Brenwald N, Reading N, Oppenheim B. Evaluation of diagnostictests for Clostridium difficile infection. J Clin Microbiol 2010;48:606e8.

[62] Tenover FC, Novak-Weekley S, Woods CW, Peterson LR, Davis T,Schreckenberger P, et al. Impact of strain type on detection of toxigenicClostridium difficile: comparison of molecular diagnostic and enzyme immu-noassay approaches. J Clin Microbiol 2010;48:3719e24.

[63] Terhes G, Urban E, Soki J, Nacsa E, Nagy E. Comparison of a rapid molecularmethod, the BD GeneOhm Cdiff assay, to the most frequently used laboratorytests for detection of toxin-producing Clostridium difficile in diarrheal feces.J Clin Microbiol 2009;47:3478e81.
































































































































































































































[64] Ticehurst JR, Aird DZ, Dam LM, Borek AP, Hargrove JT, Carroll KC. Effectivedetection of toxigenic Clostridium difficile by a two-step algorithm includingtests for antigen and cytotoxin. J Clin Microbiol 2006;44:1145e9.

[65] Turgeon DK, Novicki TJ, Quick J, Carlson L, Miller P, Ulness B, et al. Six rapidtests for direct detection of Clostridium difficile and its toxins in fecal samplescompared with the fibroblast cytotoxicity assay. J Clin Microbiol 2003;41:667e70.

[66] Van den Berg RJ, Bruijnesteijn van Coppenraet LS, Gerritsen HJ, Endtz HP, vand V, Kuijper EJ. Prospective multicenter evaluation of a new immunoassay andreal-time PCR for rapid diagnosis of Clostridium difficileeassociated diarrhea inhospitalized patients. J Clin Microbiol 2005;43:5338e40.

[67] Van den Berg RJ, Vaessen N, Endtz HP, Schulin T, van d V, Kuijper EJ. Evalu-ation of real-time PCR and conventional diagnostic methods for the detectionof Clostridium difficileeassociated diarrhoea in a prospective multicentrestudy. J Med Microbiol 2007;56(Pt 1):36e42.

[68] Vanpoucke H, De BT, Claeys G, Vaneechoutte M, Verschraegen G. Evalu-ation of six commercial assays for the rapid detection of Clostridiumdifficile toxin and/or antigen in stool specimens. Clin Microbiol Infect2001;7:55e64.

[69] Viala C, Le MA, Maataoui N, Rousseau C, Collignon A, Poilane I. Comparison ofcommercial molecular assays for toxigenic Clostridium difficile detection instools: BD GeneOhm Cdiff, XPert C. difficile and Illumigene C. difficile.J Microbiol Methods 2012;90:83e5.

[70] Walkty A, Lagace-Wiens PR, Manickam K, Adam H, Pieroni P, Hoban D, et al.Evaluation of an algorithmic approach in comparison with the Illumigeneassay for laboratory diagnosis of Clostridium difficile infection. J Clin Microbiol2013;51:1152e7.

[71] Wren MW, Kinson R, Sivapalan M, Shemko M, Shetty NR. Detection of Clos-tridium difficile infection: a suggested laboratory diagnostic algorithm. Br JBiomed Sci 2009;66:175e9.

[72] Zheng L, Keller SF, Lyerly DM, Carman RJ, Genheimer CW, Gleaves CA, et al.Multicenter evaluation of a new screening test that detects Clostridium difficilein fecal specimens. J Clin Microbiol 2004;42:3837e40.

[73] Davies KA, Longshaw CM, Davis GL, Bouza E, Barbut F, Barna Z, et al. Under-diagnosis of Clostridium difficile across Europe: the European, multicentre,prospective, biannual, point-prevalence study of Clostridium difficile infectionin hospitalised patients with diarrhoea (EUCLID). Lancet Infect Dis 2014;14:1208e19.

[74] Barbut F, Surgers L, Eckert C, Visseaux B, Cuingnet M, Mesquita C, et al. Does arapid diagnosis of Clostridium difficile infection impact on quality of patientmanagement? Clin Microbiol Infect 2014;20:136e44.

[75] Goorhuis A, Debast SB, Dutilh JC, van Kinschot CM, Harmanus C,Cannegieter SC, et al. Type-specific risk factors and outcome in an outbreakwith 2 different Clostridium difficile types simultaneously in 1 hospital. ClinInfect Dis 2011;53:860e9.

[76] Curry SR, Muto CA, Schlackman JL, Pasculle AW, Shutt KA, Marsh JW, et al. Useof multilocus variable number of tandem repeats analysis genotyping to


determine the role of asymptomatic carriers in Clostridium difficile trans-mission. Clin Infect Dis 2013;57:1094e102.

[77] Eyre DW, Griffiths D, Vaughan A, Golubchik T, Acharya M, O'Connor L, et al.Asymptomatic Clostridium difficile colonisation and onward transmission.PLoS One 2013;8:e78445.

[78] Aichinger E, Schleck CD, Harmsen WS, Nyre LM, Patel R. Nonutility of repeatlaboratory testing for detection of Clostridium difficile by use of PCR or enzymeimmunoassay. J Clin Microbiol 2008;46:3795e7.

[79] Debast SB, van KE, Oskam KM, van den Berg T, Van den Berg RJ, Kuijper EJ.Effect on diagnostic yield of repeated stool testing during outbreaks of Clos-tridium difficileeassociated disease. Clin Microbiol Infect 2008;14:622e4.

[80] Sethi AK, Al-Nassir WN, Nerandzic MM, Bobulsky GS, Donskey CJ. Persistenceof skin contamination and environmental shedding of Clostridium difficileduring and after treatment of C. difficile infection. Infect Control Hosp Epi-demiol 2010;31:21e7.

[81] Wenisch C, Parschalk B, Hasenhundl M, Hirschl AM, Graninger W. Comparisonof vancomycin, teicoplanin, metronidazole, and fusidic acid for the treatmentof Clostridium difficileeassociated diarrhea. Clin Infect Dis 1996;22:813e8.

[82] Hensgens MP, Dekkers OM, Demeulemeester A, Buiting AG, Bloembergen P,van Benthem BH, et al. Diarrhoea in general practice: when should a Clos-tridium difficile infection be considered? Results of a nested caseecontrolstudy. Clin Microbiol Infect 2014;20:O1067e74.

[83] Bauer MP, Veenendaal D, Verhoef L, Bloembergen P, van Dissel JT, Kuijper EJ.Clinical and microbiological characteristics of community-onset Clostridiumdifficile infection in the Netherlands. Clin Microbiol Infect 2009;15:1087e92.

[84] Wilcox MH, Mooney L, Bendall R, Settle CD, Fawley WN. A caseecontrol studyof community-associated Clostridium difficile infection. J Antimicrob Chemo-ther 2008;62:388e96.

[85] Alcala L, Martin A, Marin M, Sanchez-Somolinos M, Catalan P, Pelaez T, et al.The undiagnosed cases of Clostridium difficile infection in a whole nation:where is the problem? Clin Microbiol Infect 2012;18:E204e13.

[86] Reigadas E, Alcala L, Marin M, Burillo A, Munoz P, Bouza E. Missed diagnosis ofClostridium difficile infection; a prospective evaluation of unselected stoolsamples. J Infect 2015;70:264e72.

[87] Bryant K, McDonald LC. Clostridium difficile infections in children. PediatrInfect Dis J 2009;28:145e6.

[88] Wendt JM, Cohen JA, Mu Y, Dumyati GK, Dunn JR, Holzbauer SM, et al. Clos-tridium difficile infection among children across diverse US geographic loca-tions. Pediatrics 2014;133:651e8.

[89] Rogers DS, Kundrapu S, Sunkesula VC, Donskey CJ. Comparison of perirectalversus rectal swabs for detection of asymptomatic carriers of toxigenic Clos-tridium difficile. J Clin Microbiol 2013;51:3421e2.

[90] McFarland LV, Coyle MB, Kremer WH, Stamm WE. Rectal swab cultures forClostridium difficile surveillance studies. J Clin Microbiol 1987;25:2241e2.

[91] Kundrapu S, Sunkesula VC, Jury LA, Sethi AK, Donskey CJ. Utility of perirectalswab specimens for diagnosis of Clostridium difficile infection. Clin Infect Dis2012;55:1527e30.

































































































































Documents

Clinical Microbiology and Infectioninfettive.com/wp-content/uploads/2016/08/Diagnosi-C-difficile-ESCMI… · CDI testing. This guideline is intended for use among medical microbiologists,